Where it's at

I've got an idea for new directory structure layout. Ran it passed Luke and he thought it was better than the current coming together of model, tests and coded Exact Solutions. We both thought it would be wise to run it passed you before we change things up.

See http://csd.vpac.org/trac/StgNumerics/ticket/1 and http://csd.vpac.org/trac/StgNumerics/wiki/OptimalSystemTestingLayout for our efforts to make this a general setup for StgFEM and PICellerator.

* Issue to think about. What's the difference between Analytic Solution, model regression tests, Benchmarks, unit regression tests? How should our code represent the differences.

Log of existing benchmarks

So as it stands we have a lot of analytic (sanity) checks and model regression tests, and only that one sciBenchmarks in Underworld.

Most of the analytic checks and model regressions can go into StgFEM directly, this include what was there previously and many of Mirko's analytic suite.

StgFEM

*Location*	Name	Number
Apps/ThermalConvection	VelicA + VelicB	2
Apps/TempDiff	LinearTemperatureField	9
Apps/StokesMomentumUzawa	SimpleShearAnalytic + LidDrivenIsoviscousAnalytic	12
Apps/EnergySolver	AdvDiffSteadyState1D + HomogeneousNaturalBCs	5

PICellerator

*Location*	Name	Description
Apps/BuoyancyBenchmark	AnalyticPressure?	??
Apps/LidDrivenPIC
Apps/RayleighTaylor

Underworld

Just a lot of unit tests.

ExperimentalUnderworld

All of those Velic sols i added. They live in plugins/VelicAnalyticSols/

Proposed refactors

* Move Mirko's stuff from ExperimentalUnderworld to somewhere more applicable.

* Re-do the layout of the model regressions we have so far. The directories are just too complex for a freash set of eyes to understand.

* Instead of having the coded analytic solutions dispersed in the directories put them in some central directory.

* Make sure the benchmark tests are easier to find than there current state.

* In new directory structure put emphasis on including visable documentation, i.e *.pdf and also some useful output from the model regression, i.e. how quickly it ran, mem stats.

* And get the tests running again.

Benchmarks:- Julian Giordani proposed Analytic solutions are refactored to be moved together, rather than distributed, e.g.:

./Models
   <various demo XML models>
   ./Benchmarks
      bmark1.xml
      testBmark1.sh
      ./plugins
         <all the plugin analytic soln code>
      ./data
         <any necessary extra sub-xml files, or data>