h3. Recently Completed Work:

- Refining of prototype implementation of PIO in POP. \[{color:red}Dennis{color}\]
# - Added support for writing of 'holes'.
- Partial simplification of the initDecomp interface.  Remove the need to have displace arrays in the interface and added dims array.  About half way to eventual goal. \[{color:red}Dennis{color}\]
- Phase I rearrangement of directory structure \[{color:red}Loy{color}\]
- Rearrangement of directory structure and improvement of build procedure.  \[{color:red}Loy{color}\]

h3. Ongoing Work:

# -
- Evaluate extensibility of PIO interface \[{color:red}Vertenstein, Craig, Eaton, Worley, Edwards{color}\] 
- Debugging of the MCT + {binary files,pNetCDF} files.  This problem only appears to occur for a small number of IO processes on BGL (<32)  \[{color:red}Loy, Dennis{color}\]
- Debugging of {binary, pNetCDF} files.  This problem only appears to occur for a small number of IO processes on BGL (<32) \[{color:red}Dennis{color}\]
- Support for use of serial netCDF calls.  Use MCT rearranger to redistribute data to a single processor.  Disk I/O is subsequently performed by calling the serial netCDF interface. \[{color:red}Loy{color}\]
- Support for use of serial binary calls.  Use MCT rearranger to redistribute data to a single processor.  Disk I/O is subsequently performed by calling a FORTRAN read or write.\[{color:red}Loy{color}\]
- Improve support for 1D and 3D arrays in PIO.  Create a test code that exercises 1D and 3D array writes. [ {color:red}Dennis{color}

h3. Additional Work:
# -
- Need to simplify the interface to PIO_initDecomp.  This could probably be done by using a single global segmap or GDOF approach. Code within PIO_initDecomp could subsequently translate decomposition information for use by MPI data structures or pNetCDF.
- Add support for broadcasting of output arguments for I/O performed on subset of processors.
- Improved support for binary I/O
# - Support for sequential binary 
# - addition of record numbers for direct binary


- Improve parallel performance.  Possibly through the use of MPI-IO hints, or careful use of MCT and layouts.  
- More extensive testing is need.  Currently it has only been lightly tested on BGL and AIX5.
- Support for calling of PIO from non-MPI environment.  


- General restructuring of the PIO internals.  Need to modify code such that it can be easily compiled without pNetCDF, or without MPI-IO, etc...