sieve-dev Sieve questions

[Hammond, Glenn]

Matt,

I have been analyzing the Sieve code and am impressed.  I have a couple
questions:

1.  What does PCICE stand for?

2.  It appears that the element types default to triangles (2D) and tets
(3D).  Is this consistent across the board?

3.  The mesh file format (connectivity and coordinate) appears to
support implicit connectivity (i.e. no explicit list of connections
between cells/elements, but a list of cells/elements with corner
vertex/node ids and a the coordinates of the vertices/nodes).  Is this
true?  

4.  Not that I want support for explicit connectivity..., but are there
plans to support it?  We have data from other CO2 sequestration modeling
efforts that only provides explicit connectivity and volumes/areas.
   
5.  What is the format for the boundary file in:

meshpcice.c(330): void Builder::readBoundary()

6.  Everything appears to be read and set up through proc 0.  Have you
tried collective reads across all procs?  With HDF5 one can read nproc
unique hyperslabs from within the file, but the question is whether
infrastructure exists to collectively build the topology.

7.  Is it possible to share face between procs and have ghosted cells?
Right now, you have the pflotran stuff set up to share a cell, but not
the face.  With the current DA-based formulation of pflotran, we compute
the flux calculation between ghost cells and local cells on both
processors.  Although this duplicates effort, it is in place and has not
cost us much computationally.  It would be nice to still have this
option...if we so desire, even if we can be convinced otherwise.


Glenn