Skip to content

wkliao/parallel-netcdf

Repository files navigation

Parallel netCDF (PnetCDF) is a library providing high-performance parallel I/O
while still maintaining file-format compatibility with Unidata's NetCDF,
specifically the classic CDF 1, 2, and 5 formats. Although NetCDF supports
parallel I/O starting from version 4, it is built on top of HDF5 and thus its
parallel feature requires files in HDF5 format. PnetCDF is currently the only
option for parallel I/O on files in classic formats.

Starting from version 1.1.0, PnetCDF supports the CDF-5 file format, an
extension of CDF-2, that supports more atomic data types and allows users to
define large dimensions, attributes, and variables (>2B array elements).

NetCDF gives scientific programmers a space-efficient and portable means for
storing data.  By making some small changes to the NetCDF APIs, PnetCDF can use
MPI-IO to achieve high-performance parallel I/O.

More extensive installation guides can be found in file INSTALL.  Additional
information regarding the contents of the release can be found in the
RELEASE_NOTES file in the top-level directory.  Finally, the PnetCDF project
web sites below contain information on bug fixes and software releases.
    http://trac.mcs.anl.gov/projects/parallel-netcdf
    http://cucis.ece.northwestern.edu/projects/PnetCDF


Requirements:
    - PnetCDF relies on MPI-IO.  ROMIO, the most common MPI-IO implementation,
      can make use of a prefix in front of the file name to specify the
      underlying file system and override ROMIO's auto detection. A typical
      prefix would look like "nfs:" or "pvfs2:".  Bear this in mind if you use
      a file name with ':' characters.  In some cases, ROMIO might think you
      are passing in a prefix and will complain about an unsupported file
      system type.  If that happens, add a file-system prefix to your file
      name.
    - To build PnetCDF, you will need the following additional programs:
        - autotools (autoconfig, automake, libtool, m4, etc.)
        - Optional: (for PnetCDF developers)
             - either yacc or bison
	     - either lex or flex
      These are usually part of your operating system's development tools.


Build recipes:
    The process of building and installation of PnetCDF library is usually
    simple: by just running commands "./configure", "make", and "make install".
    (If more than one set of MPI compilers are available on your system, adding
    option --with-mpi=/path/to/MPI/install to the configure command line is
    recommended.) Several configure options are also available to customize it.
    See INSTALL for more information.

    Several machine- and compiler-specific build recipes are available under
    the directory of doc.
    - README.IBM        (BlueGene systems)
    - README.SX         (NEC SX systems)
    - README.CRAY       (Cray systems)
    - README.INTEL      (Intel Compilers)
    - README.LINUX      (GNU compilers)
    - README.Fujitsu    (Fujitsu systems)
    - README.K-Computer (the K computer @RIKEN)
    - README.SGI        (SGI systems, such as Endeavour @NASA)
    - README.Ubuntu     (Ubuntu)


PnetCDF User Documents:
    C API References
      http://cucis.ece.northwestern.edu/projects/PnetCDF/doc/pnetcdf-c
    Questions & Answers
      http://cucis.ece.northwestern.edu/projects/PnetCDF/faq.html

Note on supporting large files and large variables.
    - Some Fortran APIs contain arguments of 64-bit integer data type, which
      corresponds to MPI_Offset data type in their peer C APIs. Declare those
      arguments as 'integer(kind=MPI_OFFSET_KIND)'.

    - In places where you might use NF_UNLIMITED to define an unlimited
      dimension in one of the nfmpi_ routines, use NFMPI_UNLIMITED.
      NFMPI_UNLIMITED has been defined as the proper 64-bit integer type for
      nfmpi_def_dim, whereas NF_UNLIMITED might be too small. Similarly, for
      Fortran 90, use NF90MPI_UNLIMITED instead of NF90_UNLIMITED.


Mailing List
    PnetCDF user community discusses the design and use of the library on the
    [email protected] mailing list. Anyone interested in developing
    or using PnetCDF is encouraged to join. Visit the list information page
    https://lists.mcs.anl.gov/mailman/listinfo/parallel-netcdf
    for subscription instructions. This mailing list is also for announcements,
    bug reports, and questions about PnetCDF software.

    The past discussions in the mailing list are available in the follow URL:
    http://lists.mcs.anl.gov/pipermail/parallel-netcdf/.