[Dune] RFC & RFT 02/2011: proposing new core module dune-geometry

Christoph Grüninger christoph.grueninger at iws.uni-stuttgart.de
Mon Oct 17 17:23:11 CEST 2011


Dear Dune developers and user base,
it's all over town, Jö and I prepared in a branch dune-geometry which 
contains all geometry-related stuff. We think it is now mature enough 
for your testing and commenting. Hopefully it can convince you and gets 
merged to trunk.

The status can be found found at
   http://users.dune-project.org/projects/dune-geometry/wiki/Wiki
where it will be updated. For your convenience, the current status:


== Request for testing  ==

Currently it works with dune-istl and dune-pdelab from trunk.

* dune-common:
  svn checkout 
https://svn.dune-project.org/svn/dune-common/branches/develop-r6501-introduce-dune-geometry

* dune-geometry
  svn checkout 
https://svn.dune-project.org/svn/dune-grid/branches/develop-r7678-introduce-dune-geometry

* dune-localfunctions
  svn checkout 
https://svn.dune-project.org/svn/dune-localfunctions/branches/develop-r1006-introduce-dune-geometry


== Already done in dune-geometry branch  ==

* All moved headers have compatibility headers, the give a warning of an 
old header is included and automatically include the new one, too
* implementation details as jacobian_1_0_imp.hh were moved without 
compatibility headers
* dune/common/geometrytype.hh is refered by dune/geometry/type.hh
* dune/grid/quadraturerules/* has been moved to 
dune/geometry/quadraturerules/* and ../quadraturrules.hh, too
* dune/grid/genericgeometry/* has been moved to 
dune/geometry/genericgeometry/* expect geometry.hh but with 
../genericreferenceelement.hh
* vertexorder.hh, topolyfactory.hh and mockgeometry.hh are now in 
dune/geometry/ instead of dune/grid/common/ or dune/grid/utility
* dune-common, dune-grid and dune-localfunctions (still needs dune-grid 
for make check)


== Todo for dune-geometry (reason why not done)  ==

* Virtual refinement (uses Yasp grid, an expert is needed to rewrite it 
before the move from dune-grid to dune-geometry, see FS#851)
* Separate grid and geometry parts from (Basic-/Generic-)Geometry (see 
FS#959)
* mockgeometry.hh includes illegally 
dune/grid/genericgeometry/geometry.hh (only used in make check from 
dune-localfunctions, maybe we can drop this file when changing 
geometry-stuff, see FS#957)
* genericgeometry/geometrytraits.hh uses a grid as a template parameter 
but is in dune-geometry (it is included by hybridmapping.hh which is 
included in genericgeometry.hh)
* vertexorder.hh should be split: VertexOrderByIdFactory remains to 
(moves back to) dune-grid, GeneralVertexOrder and the function 
reduceOrder() belong to dune-geometry (file is used at least in 
test-monom.cc, test-edges0.5.cc and test-pk2d.cc in dune-localfunctions' 
make check)


== Tests passed  ==

* All modules successfully passed make check for dune-common, -geometry, 
-istl, -localfunctions and -pdelab. For gruenich without any additional 
grid, for joe with some grids.
* Comparing the numerics from dune-pdelab's make check gruenich had some 
meanderings between dune with dune geometry and without it. We can not 
explain why this occurs. On joe's machine this did not occur.

So please test and comment,
Christoph

-- 
Unix is simple and coherent, but it takes a genius – or at any
rate a programmer – to understand and appreciate the simplicity.
[Dennis Ritchie, 1941-2011, created C and Unix]




More information about the Dune mailing list