todo

Spectral Element Method for wave propagation and rupture dynamics.

BUGS: no known bugs
     
SOLVER:
     + bc_abso: handle "side" internally
     + more friction laws: velocity-weakening, rate-and-state, bimaterial regularization
     + Rayleigh damping
     + attenuation Q
     + higher order ABC for SH
     + PML/other absorbing layers
     + more smoothing and interpolation options for heterogeneous velocity models
     + broad band sources: Gaussian, Morlet, Berlage
     + usual Dirichlet and Neumann conditions
     + allow sources not at node
     + noise sources
     + domain decomposition database, for MPI test and multidomain coupling
     + domain reduction method with boundary traction data
     + a few triangles, Fekete or other
     + visco-plasticity
     + damage
     + poro-elasticity
     + non-linear laws (as in Shake and other commercial packages)
     + sub-cycling for simulations with disparate time-scales 
     + non-conformal mesh with mortar elements
     + implicit dynamic solvers (test pre-conditionners, blended mass)
     + implicit/explicit partitioning near small geometrical features
     + implicit quasi-static solvers: dynamic relaxation or conjugate gradient 
     
MESHER:
[1]  + kinked and curved faults 
[2]  + add surrounding strip of elements for PML
     + automated meshing starting with a triangulation: 
       Triangle (http://www.cs.cmu.edu/~quake/triangle.html)
        -->  CQmesh (http://www.dct.ufms.br/~marcelo/cqmesh.html)
         --> QMPP (http://www.dct.ufms.br/~marcelo/qmpp.html)
          --> SEM2DPACK 
     + element quality metrics
     
REFACTORING:
[1]  + 'material property' object
     + 'domain' object
     
OPTIMIZATION:
     + store coefficients elast%a in an element-by-element basis to 
       minimize memory for rectangular elements, homogeneous regions, etc.
     + more element oriented approach...  may degrade performance
       => try first with the internal-forces
     + allow single precision
     
INPUTS/OUTPUTS/POST:
     + seismograms: input sampling rate instead of number of timesteps (interpolate)
     + interpol/subsampling in all snapshot outputs
     + gnuplot plotting scripts: pm3d, vectors
     + 1D response: integrate Fabian Bonilla's contrib
     + update the 1.x tools (1D model response)
     + Nakamura's H/V method (script)
     + engineering outputs: PGA, PGV, response spectra, etc
     + Graphical User Interface front end

DOCUMENTATION:
     + improve self-documentation script: make latex output, test RoboDoc or similar
     + emc2: detailed example of quadrangulate domains for structured regions
     
PACKAGING:
     + create a SubVersion project
     + avoid unnecessary recompilation of modules (www.theochem.uwa.edu.au/fortran/recompile)
     + add compiler_variable to Makefile