wishlist

- METIS/parMETiS, Chaco, Scotch

- Parallel LSH for Stencil Gen

- GPU LU for weights
- GPU Assemble LHS stokes

- GPU ILU for stokes
- Hyperviscosity stokes


- SCVT generator

- Heat on Ellipsoid 
    -require ellipsoid projection in CVT
    -Forcing function remove analytic du/dt leave error behind; check rate of growth
    -GPU kernel for boundary conditions
    -leverage viennacl and cusp for DM and RHS (move boundary nodes to RHS)

- Hyperbolic kernel optimizations
    - Test current vs viennacl version
    - Implement cusp version for comparison
    - Benchmark with overlapped comm and comp

- Elliptic
    - Revive annulus cvt
    - Solve with Dirichlet conditions; review robin and neumann (need linear nullspace term)
    - Test parallel GMRES solve
    - Try parallel ILU0 on elliptic problem

- Stokes
    - Hyperviscosity
    - Test ContourSVD for higher cond num on weight systs
    - GPU ILU0 (Li and Saad, 2011) 
    - Compare convergence before and after interleaving
        - No precond
        - With precond

- CUSP parallel gmres
    - CPU ILU0 
    - GPU ILU0

- FSU Spear
    - Armadillo
    - (DONE) Boost
    - (module load gnu-openmpi) FFTW3
    - (module load gnu-openmpi) VTK

====================== 

- Petsc RBF-FD 
    - Should leverage the parallel solvers and preconditioners
    - May need to modify communication

- Review multi-Coloring and block-ILU
    - Coloring nodes to split matrix into blocks that can be easily updated in parallel
    - Apply ILU to small blocks; how to handle block connectors?