Changes between version 0.1 and version 0.2

This is the list of changes made after the release of Aspect version 0.1 and before 0.2. All entries are signed with the names of the author.

  1. Changed: Some models declared parameters that indicate the names of files and that contained underscores. These have been renamed to contain dashes instead, to avoid perpetual difficulties in creating the manual.
    (Timo Heister, 2013/04/28)

  2. Fixed: cleanup the various depth_average functions. Fix that all quantities got averaged the wrong way if the mesh was adaptively refined. Additionally, the velocity_magnitude and sinking_velocity was computed incorrectly.
    (Timo Heister, 2013/04/28)

  3. Fixed: Writing in any output format other than VTU and HDF5 did not yield any output files. This is now fixed.
    (Wolfgang Bangerth, 2013/04/28)

  4. Fixed: There were many places where we indiscriminately used MPI_COMM_WORLD, rather than the communicator used for the actual simulation. This is not a problem almost all the time, except for cases where Aspect is run as part of a bigger simulation.
    (Wolfgang Bangerth, 2013/04/28)

  5. New: All parts of the code now use the new MaterialModel::evaluate(). The individual functions like viscosity() got removed. To get an old material model to work, you can either move to the new setup or derive from InterfaceCompatibility instead of Interface.
    (Timo Heister, 2013/04/28)

  6. New: Checking whether the end time has been reached is now a regular termination criterion plugin. Unlike other plugins, however, it is always active.
    (Wolfgang Bangerth, 2013/04/28)

  7. New: The following plugin subsystems are now documented: mesh refinement criteria, termination criteria, and velocity boundary conditions.
    (Wolfgang Bangerth, 2013/04/28)

  8. New: There is now a new top-level parameter, "Number of cheap Stokes solver steps" that allows choosing how many iterations the Stokes solver should attempt using the "cheap" preconditioner described in the ASPECT paper. In particular, a value of zero indicates that the first phase should be skipped altogether. This is a useful strategy for problems with strongly varying viscosity for which the cheap preconditioner does not help very much and the time used for this first phase is often wasted.
    (Wolfgang Bangerth, 2013/04/22)

  9. Fixed: Entropy exponent stabilization_alpha ("alpha") should be an integer, not a double. The recommended setting is the value 2. Note that the value 1 is not explained in the ASPECT paper, only a value of 2. The various other stabilization parameters have also been better explained and cross-referenced to the ASPECT paper.
    (Timo Heister, 2013/04/17)

  10. Fixed: When running an input file that simply forgot to provide names to certain plugins (for example, that does not list anything at all for the geometry model), the resulting error message was pretty obscure and non- informative. This has now been fixed: the error message is much more descriptive and at least says where the error may have happened and what the missing parameter is.
    (Wolfgang Bangerth, 2013/04/05)

  11. New: A postprocessor to output the sparsity pattern of the block matrix. Intended mainly for debugging purposes.
    (Ted Studley, 2012/12/21)

  12. New: a module for specifying termination criteria has been added. The first termination criteria is detecting steady state RMS velocity. This scheme also allows checkpointing the simulation right before termination.
    (Eric Heien, 2012/12/11)

  13. New: The way we do mesh refinement has been completely rewritten. This is now handled via plugins that provide different refinement strategies and that can be combined if so desired. The names of refinement criteria have also been changed: "Temperature", "Velocity", and "Density c_p temperature" have been renamed to "temperature", "velocity", and "thermal energy density". Furthermore, the effect of the old name "Normalized density and temperature" can be obtained by passing the list "density, temperature". For more detail, see the manual's section on parameters that describe the mesh refinement process.
    (Wolfgang Bangerth, 2012/12/08)

  14. Fixed: In nonlinear models, we did not recompute the matrix and preconditioner for the Stokes system in every time step. This is now fixed.
    (Wolfgang Bangerth, 2012/12/04)

  15. Changed: Write the parameter file only on the root node to avoid file system locking issues.
    (Eric Heien, 2012/11/28)

  16. Changed: Adjusted the stabilization parameters to stabilize advection benchmarks without natural diffusion.
    (Rene Gassmoeller, 2012/11/16)

  17. Changed: Unified the routines for advecting temperature and compositional fields. This includes a different scaling for the temperature stabilization that may change results.
    (Rene Gassmoeller, 2012/11/15)

  18. New: If the "Use conduction timestep" parameter is true, the timestep is calculated as the minimum of the convection and heat conduction timesteps. This is to allow models where flow is very slow relative to conduction.
    (Eric Heien, 2012/11/06)

  19. New: Aspect now catches if the output directory specified in the parameter file does not exist, rather than providing a cryptic error message upon first attempt to write to this directory.
    (Wolfgang Bangerth, 2012/10/21)

  20. New: Material models can now depend on compositional fields.
    (Juliane Dannberg, 2012/10/19)

  21. New: Introduce compositional fields, that are advected by the velocity.
    (Juliane Dannberg, 2012/10/19)

  22. New: There is now a plugin that allows using GPlates velocity models as input for the velocity boundary conditions.
    (Rene Gassmoeller, 2012/10/18)

  23. Changed: SimulatorAccess moved from aspect::Postproccessor to aspect:: and can be used in material/gravity/... models too. If needed, derive from SimulatorAccess.
    (Timo Heister, 2012/09/24)

  24. Fixed: Resume does no longer trigger additional refinement times in the past.
    (Timo Heister, 2012/09/20)

  25. New: Add structures for iterated IMPES and nonlinear solvers.
    (Markus Buerg, 2012/09/20)

  26. New: InitialCondition::Function for the temperature given in the parameter file.
    (Timo Heister, 2012/09/17)

  27. New: BoundaryTemperature::Box is more flexible because it has parameters for the values for all sides of the box.
    (Timo Heister, 2012/09/17)

  28. New: Added parameter 'Temperature solver tolerance' to control the accuracy of the temperature solver.
    (Timo Heister, 2012/09/05)

  29. New: Added MaterialModel::update() and GravityModel::update() that are called before every time step.
    (Timo Heister, 2012/09/05)

  30. New: Support for HDF5/XDMF visualization output.
    (Eric Heien, 2012/08/28)

  31. New: It is now possible to select the tolerance for linear solvers from the parameter file via the global parameter "Linear solver tolerance".
    (Wolfgang Bangerth, 2012/06/30)

  32. New: Tracer particle postprocesser can write in parallel to HDF5.
    (Eric Heien, 2012/06/08)

  33. New: Tracer particle postprocessor also outputs particle.pvd for ParaView. Tracer particle data may also be output to ASCII files for easy parsing and analysis.
    (Eric Heien, 2012/06/01)

  34. New: Aspect now writes a solution.pvd for Paraview that contains a list of the files that jointly make up the entire simulation (and not just a single time step) together with the simulation time each of the files that describe a time step correspond to.
    (Wolfgang Bangerth, 2012/05/30)

  35. New: It is now selectable how the pressure should be normalized (and if it should be normalized at all). If it should be normalized at the end of each time step, one can select if the average pressure at the top surface, or the average pressure throughout the entire domain should be set to a given value (for example to zero).
    (Wolfgang Bangerth, 2012/05/15)

  36. New: The variables we output in graphical format files are now user selectable from the input parameter file. Functions that compute something from velocity, pressure and temperature (e.g., the viscosity, strain rate, etc) are now implemented as plugins like many of the other parts of Aspect. now implemented
    (Wolfgang Bangerth, 2012/05/08)

  37. Modified: Checkpoint frequency is now a user specified parameter. Frequency may be specified as either wall time or number of time steps.
    (Eric Heien, 2012/05/04)

  38. Modified: Background writing of visualization data now uses the mkstemp function rather than forking mktemp, and will continue as normal (at lower efficiency) if temporary files cannot be created.
    (Eric Heien, 2012/05/03)

  39. New: The compressibility that functions implementing the interface aspect::MaterialModel::Interface::compressibility() need to return was previously defined incorrectly as $\frac{\partial\rho}{\partial p}$. This is not what is commonly referred to as compressibility, and the function is now supposed to return $\frac 1\rho \frac{\partial\rho}{\partial p}$ instead, following the common definition of the word.
    Note that all currently implemented compressible models already did that.
    (Wolfgang Bangerth, Timo Heister, 2012/05/03)

  40. New: The number of space dimensions in which a simulation happens is now a parameter that is set in the input parameter filer, rather than statically at compile time as before.
    (Wolfgang Bangerth, 2012/04/03)

  41. New: A postprocessor module for particles was added. This module creates a random uniform distribution of particles in the mesh and moves them based on the velocity field using an RK2 scheme.
    (Eric Heien, 2012/04/02)

generated on Tue Jan 24 2017 19:53:10 by doxygen
Hosting by IWR Universit├Ąt Heidelberg