[CIG-MC] Backward incompatibility in CitcomS v3.1

[Eh Tan]

In the release note of CitcomS v3.1, I forgot to mention that v3.1 is
slightly backward-incompatible with v3.0. The same input file will
generate slightly different results in v3.0 and v3.1.

The list of backward incompatible changes:

   1. The global mesh is changed. The coordinates of all nodes are
      shifted laterally a bit. The change is not visually discernible
      when plotting the node location, but does affect the solution.
      This change is necessary to allow tracers in global model with
      nprocx and nprocy > 2.

   2. The convergence criterion of the Stokes solver is changed. A
      single parameter "accuracy" is used to control the accuracy of the
      continuity and momentum equations. "tole_compressibility" and
      "relative_err_accuracy" are not used anymore. The div/v output
      gives the residual of the continuity equation. In the previous
      versions, div/v was proportional to the continuity residual 
      times the average size of the element. The convergence of the 
      Stokes solver is achieved if either (1) div/v is less than accuracy 
      or (2) both dv/v and dp/p are less than accuracy for two consecutive
      iterations. The default value of "accuracy" was 1e-6 and is now
      1e-4.

      In the previous versions, one of the convergence criterion was 
      that div/v is less than tole_compressibility. Since div/v was 
      proportional to the size of element, a high resolution mesh would
      have small elements and thus small div/v, even when the continuity 
      residual was large. A smaller tole_compressibility was required to 
      get accurate solution for high resolution models. A common pitfall 
      was using not-small-enough tole_compressibility and getting false 
      convergence and inaccurate solution as a result.

   3. The input parameter "mantle_temp" is moved from the
      [CitcomS.solver.param] to the [CitcomS.solver.ic] section. (This
      change doesn't affect the non-pyre input file.)


-- 
Eh Tan
Staff Scientist
Computational Infrastructure for Geodynamics
California Institute of Technology, 158-79
Pasadena, CA 91125
(626) 395-1693
http://www.geodynamics.org



-- 
Eh Tan
Staff Scientist
Computational Infrastructure for Geodynamics
California Institute of Technology, 158-79
Pasadena, CA 91125
(626) 395-1693
http://www.geodynamics.org