** -*- text -*- *************************************************************
**
** Revision history of IceT.
**
** Copyright 2003 Sandia Corporation
** Under the terms of Contract DE-AC04-94AL85000 with Sandia Corporation,
** the U.S. Government retains certain rights in this software.
**
** This source code is released under the New BSD License.
**
*****************************************************************************

Revision 2.1:
	 Added Radix-k as a single-image strategy.

	 Changed the collection method for single image compositing to use
	 the built-in gather functions of MPI.  The old method had memory
	 problems with large numbers of processes and would probably be
	 inefficient anyway.

	 Changed the single image compositing methods to work strictly with
	 compressed images.  This prevents having to pad images with
	 background only to then test that same background later during
	 compression.

	 Added interlace images option that provides a hint to the
	 compositing algorithms to try shuffling the pixels in images to
	 better load balance the active pixels during compositing.

	 Patch 1: Fixed compatibility issue with compiling against MPI
	 version 1.  Fixed build issues for Windows.

Revision 2.0: A major restructuring of the IceT code comprising the
	following changes:

	* Remove the dependence of OpenGL from the core library.  Segregate
	  all OpenGL references and linkages to its own library.

	* Formalize an IceTImage object that stores pixel data.  This
          object is capable of storing pixel data with various formats
          rather than being fixed at 32-bit integers.

	* Provide the concept of a "single-image strategy."  The
          single-image strategy performs a traditional composition of a
          single image, and is used as a subroutine in some of the
          multi-tile strategies.  Since IceT is commonly used for
          compositing single tiles anyway, the development and selection of
          single-image strategies is significant.

	* Several identifiers and functions were changed in the IceT
          interface.  Many were to implement the previous features, and
          others were simply to clarify the interface.

	* Use the Glut library for OpenGL window creation in the tests.

	* Changed exposed files to conform to the current version of the
	  IceT acronym.  (It used to be ICE-T.)

	Patch 1: Fixed minor compile issues.  Fixed a common case where the
	time to compare/blend was not measured.  Fixed a corner case where
	binary swap had negative sized image partitions.

Revision 1.0: Basically the same as 0.5.4, which has been stable for years
	now, with the addition of official documentation.

Revision 0.5: Introduced the concept of ordered composition, especially
	with respect to color blending.  Once the image layer order is
	specified by the user level application, some composition
	strategies will perform image composition in that order, using the
	Porter and Duff OVER operator correctly when color blending is
	enabled.

	Added ability to handle replicated data.  If geometry is replicated
	over several processors, IceT will break the amount of screen the
	data is replicated over and divide it amongst the processors that
	all contain the same data.

	patch 1: Some changes to compile with Mac OS X and with LAM MPI.

	patch 2: Got rid of some compiler warnings.  Change the image
	inflation to use the graphics hardware to inflate with textures
	rather than inflate the image with software and then send more data
	to the graphics card.

	patch 3: Fixed the "demo bug."

	patch 4: Miscellaneous build fixes and improvements.  This
	unfortunately includes removing the man page documents.  I am in
	the process of updating the documents, so that should be ready
	"real soon now."  (As an aside, IceT is ready for its official
	"1.0" release.  I'm just waiting to finish the documentation to do
	that.)

	Fixed an issue with the MSVC 6.0 compiler and 64 bit integers.

	Fixed a problem when rendering objects behind the viewer with a
	perspective projection matrix.

Revision 0.4: Abstracted the image formats a bit.  IceT can now perform
	a composite on just the depth buffer.  This is useful when creating
	shadow buffers.  This should also make it easier to implement
	image-only compositing for parallel volume rendering.

	Abstracted the communications layer such that it is no longer
	dependent on MPI.  Of course, using MPI is still supported and is
	the only communications layer currently provided (although the user
	can potentially make his own).

	Removed icetInit, which only called icetCreateContext to make a
	"default" context.  The only real consequence of this change is
	that IceT can now be placed in a state where it has no current
	context.  This could lead to seg-faults if the user does not set
	another context as current and then tries to use IceT, but this is
	an erroneous use of IceT anyway.

	Changed build system to use CMake instead of autoconf.

	The library has been split into three parts: libicet, libicet_mpi,
	and libicet_strategies (mostly for convenience with the CMake build
	processes).  The major consequence (besides obvious linker flags)
	is that a default strategy is no longer set.

	patch 1: More updates to the build system to actually allow other
	CMake-built programs use IceT.

	patch 2: Fixed a bug with the background color.

Revision 0.3: Added the ability to render tiles smaller then the actual
	display window and then inflate the image back onto the tiles.
	Also changed the convention of how the projection matrix is set
	from using set functions or callbacks to simply setting it before
	calling icetDrawFrame.  The projection matrix is restored before
	icetDrawFrame returns.

	Removed clunky icetDisplayNodes function.  Display nodes are now
	specified in icetAddTile.

	Problems that appeared in revision 0.2 seem to be mostly fixed
	now.  Added a check in the uncompress and compare compressed
	functions to make sure there are no buffer overruns.  Adds a bit of
	overhead, but seems to have stopped an occasional seg fault that
	probably occurred when MPI dorked the data.

Revision 0.2: A whole host of fixes.  Added a regression testing suite.
	Current version seems quite stable except for intermittent problems
	with split strategy under Myrinet.  Probably either race condition
	or Myrinet/MPI bug.

Revision 0.1: Semi-stable after movement from MTIC to IceT and building
	autoconf scripts for use on Win32 rather than MSVC project files.
	Found some issues with split and reduce strategies after moving to
	Myrinet.
