CHANGE_LOG

v1.0.1 2009/05/03)
------------------
* BUG FIXES:
  - Fixed minor typo in INSTALL file. 
  - Fixed function declaration bug in lsm_geometry2d.h.

v1.0.0 (2009/02/15)
-------------------
* ENHANCEMENTS:
  - Improved support for localization/narrow-band calculations
  - Improved reinitialization via Russo & Smereka subcell fix.
  - Added support for second-order accurate Fast Marching Method 
    calculations:
    * computation of signed distance function
    * solution of Eikonal equation
  - Added support for higher-order delta functions. 
  - Added support for building single precision version of library 
    (via --enable-float configure option).
* MINOR CHANGES:
  - Updated copyright information throughout library.
  - Added many useful helper functions (e.g. curvature calculations, 
    higher-order delta function approximations, etc.).
  - Added support for gfortran compiler (default Fortran compiler) in 
    configuration scripts.
  - Redesigned Fast Marching Method implementation to use generic
    programming ideas.  The field extension, distance function calculation,
    and solution of Eikonal equation are now all "templated" on the
    number of dimensions.
* BUG FIXES:
  - Fixed the following bugs in findLineInTetrahedron():
    * for small tetrahedra, line segment returned occasionally
      extend outside of the specified tetrahedron
    * occationally the intersection point of the line with the
      tetrahedron would be missed if the intersection point lies
      on an edge of the tetrahedron.
  - Fixed several warnings


v0.9.0 (2006/12/06)
-------------------
* ENHANCEMENTS:
  - Added support for building the library outside of the source tree.
  - Added installation of documentation during install process.
  - Added support for localization/narrow-band method
  - Fast Marching Method:
    * added support for solution of general Eikonal equations
    * added support for setting grid points as being outside of domain
    * modified FMM_Core and FMM_Heap to be able to support higher 
      dimensional problems
  - Improved API for Serial LSMLIB Package.
    * Added support for masking regions as being outside of 
      computational domain
  - LSMLIB MATLAB Package
    * Added support for reinitialization
    * Modified entire package so that order of data arrays is the one
      generated by the MATLAB meshgrid() function.
  - LSMLIB Parallel Package
    * Redesigned data usage to reduce memory requirements.
    * Level set function data associated with the PatchData handles
      returned by getPhiPatchDataHandle() and getPsiPatchDataHandle() 
      methods now possesses filled ghostcell data. 
  - Boundary conditions
    * Modified parallel and serial API to be more flexible and support 
      imposing different boundary conditions for different boundary
      locations.
    * Added support for imposing homogeneous Neumann boundary conditions
      for higher order discretizations of spatial derivatives
  - New documentation on LSMLIB and level set methods available in the
    doc directory.
* MINOR MODIFICATIONS:
  - Changed getIntegratorStep() to numIntegrationStepsTaken() to avoid
    the ambiguity about whether the integrator step returned is associated
    with the beginning or end of a time step.
  - Reorganized library archives to remove circular dependencies.
* BUG FIXES:
  - Fixed createGridSetGridDims() so that it correctly sets the total
    number of grid points in the computational grid.
  - Fixed various bugs in example code.
  - Fixed configure script so that it correctly generates LSMLIB_config.h
  - Fixed configure script so that it correctly handles optimize vs. 
    debug mode.
  - Fixed inclusion of inline source code for Parallel LSMLIB Package.
  - Fixed bug in LevelSetMethodToolbox::computeStableAdvectionDt() where
    incorrect PatchData handle was used for control volume data.
  - Fixed Makefile.config.MATLAB.in so that it correctly sets the directory
    where the LSMLIB MATLAB Package should be installed.


v0.8.0 (2006/06/02)
-------------------
* ENHANCEMENTS:
  - New support for serial calculations has been added thanks to generous
    contributions by Dr. Masa Prodanovic. 
  - Modified fast marching method algorithm to set ghostcell values by
    taking the value of the nearest interior grid point rather than
    using a default value of zero.  This change was made to avoid the
    introduction of spurious zero level sets.
  - Changed logic for computation of a stable time step to improve performance.
  - MATLAB SUPPORT:  
    * added support for computation of plus/minus HJ ENO/WENO spatial 
      derivatives, TVD Runge-Kutta time integration, advancing level set 
      functions in time using the level set evolution equation.
    * added MEX-file support for Linux, Windows, and Solaris platforms.
  - SAMRAI is no longer a requirement for LSMLIB.  The serial and MATLAB 
    capabilities of LSMLIB are still available when LSMLIB is not configured
    with SAMRAI.
  - The FieldExtensionAlgorithm now supports anti-periodic boundary 
    conditions for the field variable that is being extended off of the
    zero level set.
  - The user's guide has been expanded (but will continue to grow in 
    the future releases).
  - The configure script now builds object code that is position independent
    when the architecture supports it.  This makes it possible to link
    LSMLIB into shared objects.
* API CHANGES:
  - The reinitializeLevelSetFunctions() method for the LevelSetMethodAlgorithm,
    LevelSetFunctionIntegratorStrategy, and LevelSetFunctionIntegrator classes
    has been changed to take an optional argument that specifies which 
    level set function (i.e. PHI or PSI) to reinitialize.  By default, 
    PHI is reinitialized, so codimension-one problems will not need to 
    pass an argument.
  - The FieldExtensionAlgorithm constructors now have an optional argument
    that may be used to specify that the extension field should use 
    anti-periodic boundary conditions.
  - The LevelSetMethodAlgorithm, ReinitializationAlgorithm,
    OrthogonalizationAlgorithm, FieldExtensionAlgorithm, 
    LevelSetFunctionIntegratorStrategy, and LevelSetFunctionIntegrator
    have all been modified so that reinitialization, orthogonalization,
    and computation of the extension field all take an optional max_iterations
    argument that can be used to override the stopping criterion specified
    in the input file.
  - The LevelSetMethodAlgorithm, LevelSetFunctionIntegrator, and 
    LevelSetFunctionIntegratorStrategy classes have been expanded to 
    include accessor methods for the reinitialization and orthogonalization
    interval.  These methods are provided to allow software built using
    LSMLIB to override values specified via the input file.
  - The fast marching method toolbox functions have been modified so that
    it is NOT necessary to pass ghostbox/ghostcell width information.  At
    grid points that border the edge of the computational domain, the values 
    of the distance function and extension fields are computed using only 
    the values at interior grid points.  This choice could potentially lead
    to some errors near the boundaries of the computational domain but should 
    not be a big issue for problems where the zero level set is sufficiently
    far from the boundaries. 
* MINOR IMPROVEMENTS:
  - Improved AntiPeriodicBoundaryConditionModule to be more robust about 
    detecting anti-periodic boundary conditions.
  - Improved the autoconf set_build_mode.m4 macro.
  - Added extern "C" guards around C header files.
  - Improved pattern rules in Makefile.config.in to avoid conflicts with
    compiler flags set by other software libraries.
* BUG FIXES/INTERNAL SOFTWARE MODIFICATIONS:
  - Fixed bug in OrthogonalizationAlgorithm where the wrong data is used
    to compute the gradient of fixed field during the orthogonalization
    procedure.
  - Fixed bugs in numerical kernels for field extension so that they no 
    longer allow grid cells immediately adjacent (along the coordinate
    axes) to the zero level set to be modified during the field extension 
    algorithm.
  - Redesigned OrthogonalizationAlgorithm to internally use 
    FieldExtensionAlgorithm instead of having redundant code.
  - Fixed bug in fast marching method code that caused the front to be 
    incorrectly initialized when grid points lie exactly on the zero level
    set.
  - Reorganized the logic for orthogonalization and reinitialization of
    level set functions.
  - Updated MATLAB MEX files to reflect changes in fast marching method
    and spatial derivative function signatures.
  - Fixed a minor pointer to integer conversion bug in FMM_Heap that
    arises on 64-bit architectures.
  - Fixed some bugs in the configure script.


v0.7.2 (2006/03/11)
-------------------
* ADDED FEATURES:
  - The library now uses the autoconf system for building the library.
  - Modified LSM3D_findLineInTetrahedron() so that it returns the endpoints
    of the line segment {phi=0,psi=0} oriented in the same direction as
    grad(phi) cross grad(psi)
* REMOVED FEATURES:
  - Removed ChomboVisDataWriter class since it is easier to use the VisIt
    data writer capabilities provided by SAMRAI.
* BUG FIXES:
  - Modified computation of unit normals so that they actually have norm 1 
    to machine precision (as opposed to having O(dx) error).
  - Modified computation of stable time step to avoid division by zero if
    the velocity is identically zero throughout the computational domain.
  - GCC 4.0 compatibility
    * Added explicit qualifier for the SAMRAI tbox::MPI struct.
    * Fixed syntax bug in ReinitializationAlgorithm, FieldExtensionAlgorithm
      OrthogonalizationAlgorithm


v0.7.1 (2006/02/10)
-------------------
* ADDED FEATURES:
  - Added reinitializeLevelSetFunctions() and orthogonalizeLevelSetFunctions()
    to the LevelSetMethodAlgorithm class so that users can manually
    reinitialize and orthogonalize the level set functions.
  - Changed default stopping condition for ReinitializationAlgorithm,
    FieldExtensionAlgorithm, and OrthogonalizationAlgorithm to use
    the stop distance criterion with the stop distance set to be the
    length of the largest dimension of the computational domain.
  - Improved verbose mode status output by ReinitializationAlgorithm,
    FieldExtensionAlgorithm, and OrthogonalizationAlgorithm. 
  - Minor improvements to documentation
* API CHANGES:
  - Renamed LevelSetMethodToolbox<>::copyDataToScratchSpace() to 
    copySAMRAIData().
  - Reordered arguments to constructors of ReinitializationAlgorithm,
    FieldExtensionAlgorithm, and OrthogonalizationAlgorithm because
    iteration_stop_tolerance is not a very useful stopping criterion.
    NOTE: this stopping criterion may be removed in the future...
* INTERNAL SOFTWARE MODIFICATIONS:
  - Moved LEVEL_SET_FCN_TYPE and SPATIAL_DERIVATIVE_TYPE outside of the 
    LevelSetMethodToolbox class definition to clean up the code.  They
    are now referred to by using the LSMLIB namespace (e.g. LSMLIB::ENO). 
* BUG FIXES:
  - Fixed arguments passed to ReinitializationAlgorithm and 
    OrthogonalizationAlgorithm in the LevelSetFunctionIntegrator class.
  - Removed LDFLAG* variables from Makefiles for examples since they were
    empty and could potentially override LDFLAG* variables in other included
    Makefiles.
  - Removed some C++ style comments and variable declarations in FMM_* code
    that cause errors on some compilers.


v0.7.0 (2006/02/01)
-------------------
* ADDED FEATURES:
  - Added methods in LevelSetMethodAlgorithm to create FieldExtensionAlgorithm
    objects.
  - Added accessor method for the TVD Runge-Kutta order to the 
    LevelSetMethodAlgorithm and LevelSetFunctionIntegrator classes.
  - Added cfl_number as an input parameter for the FieldExtensionAlgorithm, 
    ReinitializationAlgorithm, and OrthogonalizationAlgorithm.
  - Modified FieldExtensionAlgorithm so that the user can specify
    which component of level set function to use in field extension 
    calculation.
  - Improved design and performance of LevelSetFunctionIntegrator, 
    ReinitializationAlgorithm, and OrthogonalizationAlgorithm  by 
    removing some redundant PatchData and communications.  The extra 
    communication only occurs during vector level set calculations.
  - Added some debugging output to LevelSetFunctionIntegrator class
    which is activated when verbose_mode is set to true.
  - Redesigned LevelSetMethodAlgorithm to limit the requirements on 
    the LevelSetFunctionIntegratorStrategy class.
  - Added BoundaryConditionToolbox class to make it easier to impose
    common boundary conditions on level set functions.  Support is 
    currently provided for periodic/anti-periodic boundary conditions.
  - Added anti-periodic boundary conditions to LevelSetFunctionIntegrator,
    ReinitializationAlgorithm, OrthogonalizationAlgorithm, and
    FieldExtensionAlgorithm so that they work correctly at periodic
    boundaries (i.e. do not introduce artificial zero level set into
    solution).
  - Improved doxygen documentation that groups public methods together
    by functionality
  - Ported to BlueGene/L with XL compilers.
* INTERNAL SOFTWARE MODIFICATIONS
  - Added LSM3D_ prefix to findLineInTetrahedron() function.
    which is activated when verbose_mode is set to true.
  - Added configuration for different versions of HDF5.
  - Added use of explicit qualifiers for namespaces in explicit template
    instantiation to avoid warnings on some compilers.
  - Improved performance of fortran routines by replacing division with
    multiplication by inverse when possible.  This optimization may not be
    useful for all platforms or compiler option settings.
* BUG FIXES:
  - Fixed bugs in LevelSetFunctionIntegrator when using vector level sets.
  - Fixed bug in computation of stable time step when the external velocity
    field depends on the level set functions.
  - Fixed bugs in FieldExtensionAlgorithm, ReinitializationAlgorithm, and
    OrthogonalizationAlgorithm classes when stop tolerance is not
    specified.
  - Fixed bugs in computation of ENO3 and WENO5 spatial derivatives in 3d.
  - Fixed possible floating-point exception in 
    lsm*dComputeStableNormalVelDt() functions caused by division by zero
    when |grad(phi)| is near zero.


v0.6.3 (2006/01/17)
-------------------
* BUG FIX:  computation of volume and surface integrals now correctly returns
            the sum reduction across processors.


v0.6.2 (2006/01/13)
-------------------
* Removed unneeded methods from FieldExtensionAlgorithm, 
  ReinitializationAlgorith, and OrthogonalizationAlgorithm.
* Fixed error reporting in classes in the parallel package. 
* Added resetHierarchyConfiguration() method to LevelSetMethodAlgorithm
  class.
* Changed default spatial derivative order to depend on the spatial 
  derivative type for the LevelSetFunctionIntegrator, 
  ReinitializationAlgorithm, OrthogonalizationAlgorithm, and 
  FieldExtensionAlgorithm classes.
* Changed LevelSetFunctionIntegrator::computeStableDt() so that it 
  ignores user-specified dt values when they are zero or negative.
* Fixed several bugs throughout the library.


v0.6.1 (2006/01/09)
-------------------
* Added 2nd- and 4th-order central approximations for spatial derivatives.
* Added some const qualifiers to the classes in the parallel package.


v0.6.0 (2006/01/02)
-------------------
* Reorganized C++ interface so that all classes are in the LSMLIB
  namespace.
* Added support for vector level set method computations.
* Added support for extending fields off of the zero level set via 
  the field extension equation (see Osher & Fedkiw).
* Moved the reinitialization and orthogonalization algorithms to their
  own classes (rather than have them embedded within the 
  LevelSetMethodIntegrator class).
* Renamed methods in the LevelSetMethodPatchStrategy to make their 
  functionality more apparent.
* Added several accessor methods to LevelSetMethodAlgorithm, 
  LevelSetMethodIntegratorStrategy, and LevelSetMethodIntegrator:
  - spatial derivative type and order used in a level set method calculation
  - PatchData handle for control volume data
* Added utility functions to compute integrals over the interior and surface
  of the zero level set of the level set function.
* Removed pointer to GridGeometry from the argument list for the
  LevelSetMethodAlgorithm and LevelSetMethodIntegrator classes because
  it is redundant.
* Added component arguments (with default value 0) to several 
  LevelSetMethodToolbox methods so that they can be used on PatchData 
  with multiple components.
* Added support for doxygen source code documentation generation.
* Minor performance enhancements in numerical kernels (e.g. minimizing
  divisions, etc.)
* Minor "safety" improvements to the C++ classes. 


v0.5.3 (2005/10/25)
-------------------
* Redesigned the LevelSetMethod classes to make level set function data 
  available to the LevelSetMethodVelocityFieldStrategy class when 
  computing and initializing the velocity field.
* Added 2d example for LevelSetMethod classes when simulating motion under 
  a normal velocity.
* Minor bug fixes and corrections in comments/documentation.
* Split the samrai examples into two separate types: advection and normal 
  velocity motion.
* Added mex files to matlab package
  - 3D upwind-HJ spatial derivative calculations
  - 3D fast marching method calculations


v0.5.2 (2005/10/20)
-------------------
* Added support for motion resulting from specification of the normal 
  velocity.
  - This changed the LevelSetMethodIntegrator, 
    LevelSetMethodVelocityFieldStrategy and LevelSetMethodToolBox classes.
* Added 3d example for LevelSetMethod classes.


v0.5.1 (2005/10/12)
-------------------
* Added restart capabilities for the LevelSetMethod classes.
* Added getIntegratorStep() method for LevelSetMethodAlgorithm, 
  LevelSetMethodIntegratorStrategy, and LevelSetMethodIntegrator
  classes to keep track of how many integration time steps have been 
  taken.
* Added setCurrentTime() method to the LevelSetMethodVelocityFieldStrategy
  to make it possible to synchronize the simulation time of the velocity 
  field strategy with the simulation time of the LevelSetMethodIntegrator.
* Fixed various bugs in the code (primarily parallel bugs).


v0.5 (2005/10/11)
-----------------
* Initial full featured version of parallel, level set method library 
  for codimension-one problems.