1. 09 Dec, 2016 1 commit
  2. 08 Dec, 2016 4 commits
    • Chris Cantwell's avatar
      Merge branch 'fix/dont_overwrite_chk' into 'master' · 41212019
      Chris Cantwell authored
      dont overwrite chk files if the solver was restarted
      in case f a restart, read the number from the restart file and use this as starting point for the next chk files, so that when restarting from foo_215.chk, the first chk file written is not foo_0.chk but foo_215_0.chk (which contains the initial conditions and should in most cases be identical to foo_215.chk). The next file will be foo_216.chk with the new timestep.
      In case foo_215_0.chk already exists, we write foo_215_1.chk and so on.
      The filters still overwrite everything, maybe we should change that too?
      See merge request !707
    • Dave Moxey's avatar
      Merge branch 'fix/linearise_mapping' into 'master' · fd5eb756
      Dave Moxey authored
      This MR fixes the linearise threshold routine which was being far too over aggressive in the linearisation of prism elements.
      This is because it was measuring linear deformation when considering the quality, this has now been fixed.
      See merge request !717
    • Dave Moxey's avatar
      Merge branch 'master' into 'fix/linearise_mapping' · f7866e7c
      Dave Moxey authored
      # Conflicts:
      #   utilities/NekMesh/ProcessModules/ProcessLinear.cpp
    • Dave Moxey's avatar
      Merge branch 'feature/helmsmooth3D' into 'master' · 200fd1d0
      Dave Moxey authored
      This MR enables the hemholtz smoother for 3D problems
      It also activates PETSc MUMPS support when a Fortran compiler is present because currently, only direct solvers can reliably handle the resulting linear system.
      See merge request !714
  3. 07 Dec, 2016 6 commits
  4. 06 Dec, 2016 4 commits
  5. 05 Dec, 2016 12 commits
  6. 03 Dec, 2016 2 commits
  7. 02 Dec, 2016 3 commits
  8. 01 Dec, 2016 4 commits
  9. 30 Nov, 2016 3 commits
  10. 25 Nov, 2016 1 commit
    • Dave Moxey's avatar
      Merge branch 'feature/Compressible-opt' into 'master' · e52b1536
      Dave Moxey authored
      Tidy compressible solver
      This MR makes several changes which I believe make the code for the compressible solver a lot more organised. I originally intended to work on the efficiency of the solver, but I only made a few changes towards this. The main changes are:
      - Move boundary conditions to separate class using factory
      - Move functions for converting variables (e.g. GetPressure) to a separate auxiliary class
      - Remove the ProblemType option and have RinglebFlow and IsentropicVortex as separate equation systems derived from Euler
      - Try to be more consistent about what should  be in the CompressibleFlowSystem base class and what is specific to the particular equations
      - Rewrite the GetFluxVector functions in a more compact form, which depends on the problem dimension only in the loop limits
      - Allow passing Fwd and Bwd trace space to Advection and Diffusion operators as optional parameters. If these parameters are not set they are calculated as usual, but using them helps in saving some communication cost by reusing the values passed to the advection in the diffusion (perhaps a better condition is required to determine when to precompute these values).
      I also implemented a semi-implicit DG/HDG for the unsteady advection diffusion, which seems to be working fine.
      See merge request !684