1. 15 Dec, 2014 2 commits
  2. 14 Dec, 2014 3 commits
  3. 13 Dec, 2014 8 commits
    • Dave Moxey's avatar
      Merge branch 'fix/manifold-history-points' into 'master' · 1b4cb847
      Dave Moxey authored
      Fix/manifold history points
      This MR fixes (again) the history points for the manifold case. Since the points may not truly lie in the manifold, we have to be forgiving. The algorithm now proceeds as follows
      - Each process searches for the closest element in its partition which might contain the history point.
      - Distances are recorded and the minimum is found across all processes to decide which process should 'own' the history point.
      - In the case where multiple processes satisfy the above (e.g. point lies on partition boundary), the one of highest rank is chosen.
      A warning is printed if the distance from the point to the element is greater than the geometric tolerance constant. In this case, the history point will write out data from the projection of the point on the element.
      See merge request !410
    • Dave Moxey's avatar
      Merge branch 'feature/MoveAdvectionClass' into 'master' · bc340bb8
      Dave Moxey authored
      Feature/move advection class
      This MR concerns a restructuring of the handling of advection objects:
      - Advection terms in IncNavierStokesSolver now derive from the SolverUtils Advection base class
      - A new AdvectionSystem holds the Advection term at Driver level and sits between UnsteadySystem and a specific solvers EquationSystems (to allow Advection objects to be manipulated at Driver level)
      - Moved instantiation of the extrapolation terms to the most derived classes in IncNavierStokesSolver. Previously m_fields needed to be initialised (through a call to UnsteadySystem::v_InitObject()), the m_pressure pointer set, and then the extrapolation term initialised through calling IncNavierStokes::v_InitObject(). However, this now also calls UnsteadySystem::v_InitObject resulting in m_fields being initialised twice and m_pressure and m_fields[m_nConvectiveFields] being different.
      - Some tidying up of the advection term code.
      A similar structure can be adopted for the diffusion terms. The AdvectionSystem class therefore inherits UnsteadySystem virtually to allow this to be implemented.
      See merge request !408
    • Chris Cantwell's avatar
    • Chris Cantwell's avatar
    • Chris Cantwell's avatar
      Merge branch 'feature/refactor-assemblymaps' into 'master' · 0ba33b73
      Chris Cantwell authored
      Refactor assembly maps
      This MR refactors `AssemblyMapCG(1,2,3)D` into `AssemblyMapCG`, and removes the dimension-dependent map creation functions from `AssemblyMapDG`, replacing them with a single function. This was achieved by taking the 3D versions and adapting where necessary to make them work in 1D and 2D. There are some library changes as a consequence, but these are pretty minor:
      - `StdExpansion::GetFaceOrient` -> `StdExpansion::GetForient` (to be consistent with `StdExpansion::GetEorient`) and implementation is moved into `Expansion3D`.
      - Added `StdExpansion::GetNtrace` which returns the number of trace elements for a given expansion.
      - A few changes to `Geometry` to add `GetFid` at the base class level. Also added `GetTid` which returns geometry IDs of corresponding trace elements.
      - Definition and implementation of some of the DG routines is now in `Expansion` instead of `StdExpansion`.
      See merge request !380
    • Chris Cantwell's avatar
      Tidy up formatting of DriverSteadyState. · 9519359b
      Chris Cantwell authored
    • Chris Cantwell's avatar
    • Chris Cantwell's avatar
      Merge branch 'feature/MoveAdvectionClass' into feature/AutomatedSFD · a840ddb0
      Chris Cantwell authored
  4. 11 Dec, 2014 2 commits
  5. 10 Dec, 2014 3 commits
  6. 09 Dec, 2014 1 commit
  7. 08 Dec, 2014 1 commit
  8. 07 Dec, 2014 3 commits
  9. 06 Dec, 2014 5 commits
  10. 05 Dec, 2014 7 commits
  11. 04 Dec, 2014 1 commit
  12. 03 Dec, 2014 3 commits
  13. 02 Dec, 2014 1 commit