Permalink
Commits on Aug 26, 2016
  1. @jwpeterson

    Merge pull request #1078 from jwpeterson/bvst

    Add BoundaryVolumeSolutionTransfer class.
    jwpeterson committed on GitHub Aug 26, 2016
  2. @jwpeterson

    Run boostrap.

    jwpeterson committed Aug 26, 2016
  3. @jwpeterson

    Update Makefile.am's.

    jwpeterson committed Aug 26, 2016
  4. @jwpeterson

    Add BoundaryVolumeSolutionTransfer class.

    This class can be used for transferring solutions between the surface
    of a volume mesh and the BoundaryMesh associated with that surface.
    This is joint work with with Xikai Jiang from Argonne National
    Laboratory (@xikaij) who is the original author of the transfer.
    
    See also: X. Jiang et al., "An O(N) and parallel approach to integral
    problems by a kernel-independent fast multipole method: Application to
    polarization and magnetization of interacting particles," The Journal
    of Chemical Physics vol. 145, 064307, http://dx.doi.org/10.1063/1.4960436.
    jwpeterson committed Aug 26, 2016
  5. @roystgnr

    Merge pull request #1076 from dschwen/mpi_ad_cache_race

    Fix race condition due to concurrent MPI file writes
    roystgnr committed on GitHub Aug 26, 2016
  6. @roystgnr

    Merge pull request #1077 from pbauman/femcontext-doc-update

    Better documentation for FEMContext::<elem,side,edge>_reinit
    roystgnr committed on GitHub Aug 26, 2016
Commits on Aug 25, 2016
  1. @pbauman

    Better documentation for FEMContext::<elem,side,edge>_reinit

    pbauman committed Aug 25, 2016
  2. @jwpeterson

    Merge pull request #1075 from jwpeterson/drop_cerr

    Add support for --drop-cerr command line argument.
    jwpeterson committed on GitHub Aug 25, 2016
  3. @dschwen

    Fix race condition due to concurrent MPI file writes

    dschwen committed Aug 25, 2016
  4. @jwpeterson

    Add support for --drop-cerr command line argument.

    There was also a libmesh-devel thread discussing this: http://tinyurl.com/hseqsm4
    jwpeterson committed Aug 25, 2016
  5. @jwpeterson

    Merge pull request #1074 from jwpeterson/remove_throw

    Remove libmesh_parallel_only() from libMeshInit destructor.
    jwpeterson committed on GitHub Aug 25, 2016
  6. @roystgnr

    Merge pull request #1069 from pbauman/time-solver-fixes

    FEMSystem TimeSolver Fixes
    roystgnr committed on GitHub Aug 25, 2016
  7. @pbauman

    Add new NewmarkSolver unit test

    For this one, we should be able to integrate exactly a linear
    function (in time) of acceleration. This required both the beta
    setter (since you need beta = 1/6 for this case) and the reinit_func
    fix in the NewmarkSolver.
    pbauman committed Aug 24, 2016
  8. @pbauman

    Rework NewmarkSolverTest a bit

    This is to prepare for adding a new test where we reset beta.
    pbauman committed Aug 24, 2016
  9. @pbauman

    Add reinit_func to NewmarkSolver, call in residual

    We need to call reinit_func in order to set t = t_{n+1} for evaluating
    the residual for Newmark. Of course, we'll also need it for the mesh
    motion if it's there.
    pbauman committed Aug 24, 2016
  10. @pbauman
  11. @pbauman

    Add Euler2Solver unit test for second order

    This test exercises the reinit_func fix. Before, we didn't have
    the correct time in the context so we didn't integrate this ODE
    exactly as we should have. Now we do.
    pbauman committed Aug 24, 2016
  12. @pbauman

    Add initial reinit_func(1.) call in Euler2Solver

    We need the call to reinit_func to set the time, t, in the context
    to the correct value. Also added clarifying comments in EulerSolver
    and Euler2Solver that we're also setting the time in addition to
    possibly resetting the mesh if there's mesh motion.
    
    There is probably a way to make this more efficient such that we
    only call reinit_func twice in Euler2Solver, but I didn't put any
    thought into it.
    pbauman committed Aug 24, 2016
  13. @pbauman

    Only call elem_fe_reinit in elem_reinit if there's mesh motion

    Otherwise, it's unnecessary since elem_fe_reinit is called before
    assembling each element residual contributions.
    pbauman committed Aug 24, 2016
  14. @jwpeterson

    Remove libmesh_parallel_only() from libMeshInit destructor.

    As (correctly) pointed out by the GCC 6.1 compiler, you generally
    don't want to throw from a destructor, and this function, if it
    failed, would throw via the following chain of macro calls:
    
    ~LibMeshInit -> libmesh_parallel_only() -> libmesh_assert -> libmesh_error_msg -> LIBMESH_THROW
    
    This commit implements @roystgnr's suggestion of just removing it.
    
    Closes #1073.
    jwpeterson committed Aug 25, 2016
  15. @jwpeterson

    Merge pull request #1072 from friedmud/master

    Remove memory churn in compute_affine_map
    jwpeterson committed on GitHub Aug 25, 2016
  16. @jwpeterson

    Update citations.

    jwpeterson committed Aug 25, 2016
  17. @roystgnr

    Merge pull request #1071 from jwpeterson/metis_fix

    Don't assume interior_parents() are always part of the same Mesh.
    roystgnr committed on GitHub Aug 25, 2016
  18. @friedmud

    Remove memory churn in compute_affine_map, improving threading perfor…

    …mance
    friedmud committed Aug 25, 2016
Commits on Aug 24, 2016
  1. @jwpeterson

    Make sure that every interior neighbor is actually in the same Mesh.

    Error out if any valid interior neighbors are not found in the
    global_index_map, don't just assert in dbg mode and keep running with
    possibly wrong connectivity in opt mode.
    jwpeterson committed Aug 24, 2016
  2. @jwpeterson

    Use vectormap::find() to avoid extra lookup.

    jwpeterson committed Aug 24, 2016
  3. @jwpeterson
  4. @jwpeterson
  5. @jwpeterson
  6. @roystgnr

    Merge pull request #1068 from pbauman/scalar-time-solver-test

    TimeSolver Unit Tests
    roystgnr committed on GitHub Aug 24, 2016
  7. @pbauman

    bootstrap

    pbauman committed Aug 24, 2016
  8. @pbauman

    Add new tests to build system

    pbauman committed Aug 24, 2016
  9. @pbauman

    Add NewmarkSolver unit test

    As with the first order tests, we use a very basic FEMSystem to
    test integrating a second order ODE in time.
    pbauman committed Aug 24, 2016
  10. @pbauman

    Add unit tests for Euler(2)Solver

    This uses FEMSystem to implement simple ODEs to integrate in time
    and unit test the Euler solvers. In particular, we have one that
    should be integrated exactly for first order solvers and one
    that should be integrated exactly with second order solvers.
    
    There are currently two issues I've found here.
    1: Using a SCALAR
    variable (and nonlinear_time_derivative and nonlinear_mass_residual)
    gives the *WRONG* answer. I'm not sure why yet - this needs to be
    investigated further.
    2: Euler2Solver does not integrate exactly the LinearTimeFirstOrderODE
    System. I've found the problem and will fix it subsequently.
    pbauman committed Aug 23, 2016
Commits on Aug 23, 2016
  1. @roystgnr

    Merge pull request #1067 from permcody/issue_1062

    Add vector of vector scatter implementation
    roystgnr committed on GitHub Aug 23, 2016