[cig-commits] r16767 - short/3D/PyLith/trunk

[brad at geodynamics.org]

Author: brad
Date: 2010-05-23 09:34:51 -0700 (Sun, 23 May 2010)
New Revision: 16767

Modified:
   short/3D/PyLith/trunk/TODO
Log:
Updated TODO.

Modified: short/3D/PyLith/trunk/TODO
===================================================================
--- short/3D/PyLith/trunk/TODO	2010-05-23 04:44:02 UTC (rev 16766)
+++ short/3D/PyLith/trunk/TODO	2010-05-23 16:34:51 UTC (rev 16767)
@@ -111,25 +111,16 @@
 
   * GenMaxwellPlaneStrain (plane strain Generalize Maxwell model) [Charles]
 
-  * Better preconditioning [Matt/Brad/Charles]
-    + Need settings for Schur complement
+  * Better preconditioning [Matt/Brad]
 
-    We should temporarily hardwire the Schur complement settings (enable
-    with #define) and see how well it works.
+    Diagonal preconditioner does not work very well. It looks like the
+    preconditioner needs off-diagonal terms consistent with the
+    coupling of the Lagrange multiplier constraints (sparsity pattern
+    matches the union of the matrices associated with the fault
+    meshes).
 
-    MATT is thinking about how to setup Schur complement settings
-    (recursive specification of field splits).
-
     Need field split working for both SolverLinear and SolverNonlinear.
 
-    Preconditioner Object
-
-      Owned by formulation?
-      Solver::initialize() sets pc object in Preconditioner
-
-      use_field_split
-      use_custom_fault_pc
-
   * Optimization [Brad]
     + Specialized elasticity integrator objects for Tri3, Quad4, Hex8