2007.04.17 - Added an atomicity option.  Also, changed the input
method for hints.  Removed all nongeneric hints from the testing
interface.  Updated the README to reflect quick and easy usage.

2007.01.25 - Fixed errors for checking the right amount of bytes.
Fixed some filenames, updated README.  Noted that -D 1 will have some
problems with correctness for MPI datatype cases which have not yet
been resolved.  Please email me for more information or the solution.

2007.01.16 - Fixed an error with MPI_File_set_size() for collective
I/O with respect to the call being collective.

2007.01.04 - Fixed a calculation error with the MPI_File_set_size()
command.  File sizes are now correctly calculated.

2006.12.13 - Updated the README a bit and made changes so that HPIO
will compile with ANSI C.  Cleaned up wthe HPIO help output.

2006.08.29 - Revised some of the documentation.  Made changes so the
test compiles and works with MPICH1-1.2.7p1.  Changed the dup test to
cnt_vec.  Fixed struct to use MPI_Aint.  Added a new output file to
name all failed tests.

2006.08.24 - Added the new test 'defined' to the test suite.  Changed
a lot of input options.

2006.08.11 - Solved the reading past the end of file for MPI_UB which
is beyond the last byte.  Basically, use MPI_File_set_size to change
the EOF.  Added a new debug method to pipe output to multiple streams.

2006.08.10 - Added correctness test (human).  Now there are 3 ways to
verify.  (1) Simply check the accessed data.  (2) Check the manually
provided data.  (3) Check against the generated data.  Added a new
debuging library.  Use an environment variable DEBUG_MASK to turn on
various debugging messages.  Added output files to generate both the
displayed data as well as the excel ready data.

2006.08.01 - Changed the verification method.  Provides 2 levels of
verification.  (1) Simply check the accessed data. (2) Check the
entire file/buffer (write/read).  Also, added the human test array
(primarily for correctness testing).

2006.07.20 - Fixed the timing so that processes sort repetitions based
on the aggregate total I/O time (I/O time + synchronization time).
Added another averaging method for "best repetition," which allows the
test to run more smoothly on shared file systems.

2006.07.13 - Added the editor comments to keep things consistent to
*.c and *.h files.

2006.07.10 - A lot of new features.  First, a resume option (-R) to
restart the test at the failed position.  Another option to handle a
full generation of files (-g) before any writing/reading occur to get
rid of weird datasieving performance.  Fixed the read options to
automatically generate the proper files before doing any reading.
Added another option -k to not delete.

2006.06.30 - Bugfix for the help screen to correct the -D option for
struct and vector.  Fixed the verify mode to not require a parameter.
Fixed script for creating new versions to use push.

2006.06.28 - Added another pattern datatype option to use VECTOR or
STRUCT.  Changed the verify mode to use different scenarios to read
back the written data (i.e. individual and collective or VECTOR and
STRUCT).  Added a script for creating new versions.

2006.06.27 - Added a single test mode to look case by case.  Changed
the messages to indicate when options are being changed without the
user's knowledge.  Program now does not override any user options.

2006.06.26 - Added cb_node option to change the number of aggregators
in use.  Added some checks for string copying with sprintf and
strncpy.

2006.06.19 - Changed cb_node_list to cb_config_list (Kenin's suggestions).

2006.06.15 - Checks parameters for no repeats.

2006.05.25 - Force all collective I/O to use all processes as
aggregators even on a single machine.

2006.03.23 - Added another noncontig method for persistent file realms
(PFRs).  Added the same file parameter to show how PFRs can help
performance.

2005.09.02 - Added more test parameters output and a new method for
flushing files that should clash less often with the 2GB file limit.
Fixed the repetition system to only do things once that take longer
than 10 minutes.

2005.08.31 - Fixed the collective case to use the MPI_File_open with
MPI_COMM_WORLD.

2005.08.17 - Eliminated memory leaks of MPI Hints and directory.
Fixed the reduce process of times to use the maximum time and its
components.

2005.08.15 - Fixed the bug to directories.  Works with PVFS2 and MPICH2

2005.08.12 - Fixed the averaging process, got the results to verify
and added a method for killing the cache in the process.

2005.08.10 - Got most of the basic stuff working.  Next step is to
verify the results as well as fix the averaging process.
