I remade a benchmark and finally when using boost::thread_specific_ptr,
using a std::vector is faster than the dedicated vector-class I previously
use.
I also retry using a tuple instead of a dedicated class to store variables
useful for the emulated recursion but tuple is (on my machine) slower.
benchmark results are the following (points are randomly taken in a sphere,
time is in seconds and is the mean of 10 idendical runs)
nb points | 1000 100000 100000 1000000 10000000
-----------------------------------------------------------------------------
trunk version | 0.0088 0.0928 0.9865 10.216 102.99
with vector | 0.0096 0.0900 0.9817 10.075 102.24
with vector and tuple | 0.0100 0.0948 0.9981 10.219 104.58
recursive version | 0.0080 0.0936 0.9732 10.141 102.33
svn+ssh://lrineau@scm.gforge.inria.fr/svn/cgal/branches/CGAL-3.6-branch
........
r56835 | lrineau | 2010-06-17 12:56:52 +0200 (Thu, 17 Jun 2010) | 6 lines
Remove the constructor Gmpfr(long double) on Microsoft Visual C++. A big
comment in the source code explains why.
The testsuite will check that the construction of Gmpfr from a long double
on MSVC still works and produces the right Gmpfr.
........
r56864 | afabri | 2010-06-18 11:04:47 +0200 (Fri, 18 Jun 2010) | 1 line
Use tie from boost::
........
r56865 | afabri | 2010-06-18 11:11:49 +0200 (Fri, 18 Jun 2010) | 1 line
Use bind from boost::
........
r56866 | afabri | 2010-06-18 11:38:50 +0200 (Fri, 18 Jun 2010) | 1 line
Use bind from boost:: (detected in Mesh_3 VC10 testsuite)
........
r56867 | lrineau | 2010-06-18 11:39:24 +0200 (Fri, 18 Jun 2010) | 3 lines
cmake-2.8.2rc2 is out.
/bigobj is necessary
........
r56868 | afabri | 2010-06-18 11:52:37 +0200 (Fri, 18 Jun 2010) | 1 line
Add #include <fstream>
........
r56869 | afabri | 2010-06-18 11:55:33 +0200 (Fri, 18 Jun 2010) | 1 line
Shorten filename as with path it exceeds easily 256 letters which poor Visual C++ can't handle
........
r56870 | afabri | 2010-06-18 12:24:30 +0200 (Fri, 18 Jun 2010) | 1 line
Use tie from boost::
........
r56876 | lrineau | 2010-06-18 16:40:36 +0200 (Fri, 18 Jun 2010) | 3 lines
New try to fix the issue of Gmpfr(long double) with MSVC and libmpfr-1.dll
compiled by Mingw.
........
r56895 | lrineau | 2010-06-20 23:16:40 +0200 (Sun, 20 Jun 2010) | 3 lines
Using boost::bind is not sufficient" "bind" without qualifier was
ambiguous, according to MSVC2010, with std::bind (from C++0x).
........
r56896 | lrineau | 2010-06-20 23:18:29 +0200 (Sun, 20 Jun 2010) | 3 lines
Qualify "bind" with "boost::", to avoid the ambiguity (according to
MSVC2010), with std::bind (C++0x).
........
r56897 | lrineau | 2010-06-20 23:19:17 +0200 (Sun, 20 Jun 2010) | 2 lines
Stupid typo!
........
--create an internally used vector that can reserve its size when constructed
--create an internally used class to store information required to emulated the call stack
--the stack used to emulate the call stack is a static thread-safe vector of the function (thus a memory overhead)
These choices have been made so that the running time of the triangulation is still the same.
by namespace CGAL { and } //namespace CGAL. in all .h and .cpp files
in a directory.
Apply it to all packages in the trunk
Remove macro definition from the config.h file.
svn+ssh://lrineau@scm.gforge.inria.fr/svn/cgal/branches/CGAL-3.5-branch
Personal note: Actually all merges were already done in the trunk manually
by authors, as far as I have seen. I have ignored several conflicts in
Mesh_3 and packages from Laurent Saboret, because I know the authors of
these packages already manage patchs to 3.5 and trunk manually. All the
changes here seem to be fixes to package descriptions (change -08 to -09).
........
r51561 | stayeb | 2009-08-27 17:41:39 +0200 (Thu, 27 Aug 2009) | 8 lines
Merge r51076
From trunk/Mesh_3/include/CGAL/Labeled_image_mesh_domain_3.h to branches/CGAL-3.5-branch/Mesh_3/include/CGAL/Labeled_image_mesh_domain_3.h
r51076 | stayeb | 2009-08-05 17:40:23 +0200 (Wed, 05 Aug 2009) | 3 lines
* Fix bug with image bounding box computation
* Add internal(undocumented) usefull genericity:
+ add Wrapper template parameter (with default value) in Labeled_Image_mesh_domain_3
........
r51678 | lsaboret | 2009-09-02 14:16:10 +0200 (Wed, 02 Sep 2009) | 6 lines
svn merge bug fix from trunk to CGAL-3.5-branch:
r51677 | lsaboret
Port to Windows 64
........
r51694 | stayeb | 2009-09-02 16:30:21 +0200 (Wed, 02 Sep 2009) | 1 line
add multilabel_mesh_small picture.
........
r51699 | lrineau | 2009-09-02 16:42:46 +0200 (Wed, 02 Sep 2009) | 35 lines
Merge the following revision from trunk into the 3.5 branch:
| ------------------------------------------------------------------------
| r51688 | lrineau | 2009-09-02 16:07:12 +0200 (Wed, 02 Sep 2009) | 2 lines
|
| A new solution for the "how_to_cite stuff"
|
| ------------------------------------------------------------------------
| r51689 | lrineau | 2009-09-02 16:08:14 +0200 (Wed, 02 Sep 2009) | 2 lines
|
| Remove the last non-constant part of the script makebiblio !!! :-)
|
| ------------------------------------------------------------------------
| r51679 | lrineau | 2009-09-02 14:48:37 +0200 (Wed, 02 Sep 2009) | 4 lines
|
| Fix after Monique's suggestion... Just kidding! :-)
|
| Actually, we do not want CGAL Editorial Board to be abbrev. to "C.E.Board".
|
| ------------------------------------------------------------------------
| r51680 | lrineau | 2009-09-02 14:52:09 +0200 (Wed, 02 Sep 2009) | 2 lines
|
| Protect Minkowski and Boolean with {}.
|
| ------------------------------------------------------------------------
| r51692 | lrineau | 2009-09-02 16:15:22 +0200 (Wed, 02 Sep 2009) | 6 lines
|
| Fixes:
| - Remove the links "BACK TO INDEX" that were auto-generated by
| bibtex2html
| - Remove several "and and".
| - Replace the labels: "08" -> "09".
|
| ------------------------------------------------------------------------
........
r51702 | yvinec | 2009-09-02 16:51:07 +0200 (Wed, 02 Sep 2009) | 3 lines
fixed PkgDescription.tex
........
r51714 | spion | 2009-09-02 18:46:41 +0200 (Wed, 02 Sep 2009) | 4 lines
Backport revision 51047.
Replace wrong use of $$ by \ccc{}.
........
r51738 | lrineau | 2009-09-03 09:24:23 +0200 (Thu, 03 Sep 2009) | 2 lines
Import revision 51737 from trunk: fix the how_to_cite_cgal.html file
........
r51758 | lrineau | 2009-09-03 12:14:14 +0200 (Thu, 03 Sep 2009) | 1 line
Fix all \ccPkgHowToCiteCgal at once to turn 08 into 09
........
r51769 | lrineau | 2009-09-03 14:03:23 +0200 (Thu, 03 Sep 2009) | 14 lines
Merge from trunk:
| ------------------------------------------------------------------------
| r51754 | teillaud | 2009-09-03 11:45:05 +0200 (Thu, 03 Sep 2009) | 2 lines
|
| added missing file, needed by makebiblio
|
| ------------------------------------------------------------------------
| r51755 | teillaud | 2009-09-03 11:48:45 +0200 (Thu, 03 Sep 2009) | 2 lines
|
| layout file for makebiblio (use of ccs)
|
| ------------------------------------------------------------------------
........
r51771 | lrineau | 2009-09-03 14:20:50 +0200 (Thu, 03 Sep 2009) | 10 lines
Merge from trunk:
| ------------------------------------------------------------------------
| r51770 | teillaud | 2009-09-03 14:14:17 +0200 (Thu, 03 Sep 2009) | 3 lines
|
| a few buggy labels fixed
| obviously, it was not a good idea to write them using the keyboard alone, not even the mouse... :(
|
| ------------------------------------------------------------------------
........
r51778 | lrineau | 2009-09-03 16:38:42 +0200 (Thu, 03 Sep 2009) | 3 lines
Fill the file global_dont_submit, so that one can simplify the script that
creates releases.
........
r51787 | lsaboret | 2009-09-03 18:00:04 +0200 (Thu, 03 Sep 2009) | 15 lines
Merge bug fixes from trunk to CGAL-3.5-branch:
r51785 | lsaboret
* Fixed images width (as requested by LR):
set max-width instead of width in HTML.
* Also set width=1.0 for most images in PDF (prettier).
r51749 | lsaboret
Fixed warning with boost >= 10.38 in Point_set_processing_3.
r51683 | lsaboret
Fixed compilation error due to the global change "CGALi" -> "internal" in trunk
(this revision is merged into CGAL-3.5-branch by safety)
........
r51814 | teillaud | 2009-09-04 10:16:38 +0200 (Fri, 04 Sep 2009) | 2 lines
files generated after yesterday's fixes on manual-3.5.bib
........
r51831 | lsaboret | 2009-09-04 13:38:11 +0200 (Fri, 04 Sep 2009) | 6 lines
Merge bug fix from trunk to CGAL-3.5-branch:
r51830 | lsaboret
Fixed compilation error introduced in r51749
........
r51868 | lrineau | 2009-09-07 17:01:48 +0200 (Mon, 07 Sep 2009) | 2 lines
Fix the non-filtering of BBox_3-Segment_3 intersection.
........
r51889 | palliez | 2009-09-10 17:02:29 +0200 (Thu, 10 Sep 2009) | 1 line
fix pca bug
........
r51895 | lsaboret | 2009-09-10 18:34:35 +0200 (Thu, 10 Sep 2009) | 66 lines
Merge bug fixes from trunk to CGAL-3.5-branch:
r51894 | lsaboret
Added traces
r51893 | lsaboret
Fixed bug introduced in r51683:
(Windows) developers compiling Surface_reconstruction_points_3 with CGAL 3.5 beta 1 need to compile with Surface_mesher package from svn in order to get r51028 bug fix (orientation of reconstructed meshes).
r51887 | lsaboret
Fixed bug in test suite:
* The bug occured (sometimes) in test suite when calling
> APSS_reconstruction_test data/robocat_deci.off
=> assertion failure "CGAL::Polyhedron_incremental_builder_3<HDS>::
lookup_halfedge(): input error: facet 1462 shares a halfedge from vertex 95 to vertex 808 with facet 1312."
in output_surface_facets_to_polyhedron().
This means that the reconstructed surface is not manifold.
* The reason seems that (on this example) APSS implicit function is not smooth enough for make_surface_mesh().
* The fix was to set APSS smoothness factor to 6 (default is 2).
r51886 | lsaboret
Partial fix of assertion failure "(x,y,z) is already inserted on surface" in make_surface_mesh():
* The bug could be reproduce by:
> poisson_reconstruction Compressor_top_clean_65kpoints.pwn out.off -sm_distance 0.75
or
> APSS_reconstruction robocat_deci.off out.off -smooth 6
(try several times)
* The fix is to set Implicit_surface_3's dichotomy error to make_surface_mesh's approximation distance/1000 (instead of /10). This makes make_surface_mesh() behavior more reproductible.
* My guess is that in both implicit functions are not smooth, and that make_surface_mesh's parameters must be tuned to make them appear smooth to the algorithm.
r51884 | lsaboret
Temporary traces: turn on CGAL assertions
r51883 | lsaboret
Temporary traces: turn on CGAL assertions
r51873 | lsaboret
Fixed bug in make_surface_mesh(implicit surface, sphere, Manifold_tag) when the implicit surface crosses the sphere:
* The bug could be reproduced on Linux/g++ 4.3.1 by:
> poisson_reconstruction Compressor_top_clean_30kpoints.pwn Compressor-poisson.off
or
> APSS_reconstruction robocat_deci.off robocat_deci-apss.off
=> infinite loop that crashes when the process max memory is reached, or
assertion failure "(x,y,z) is already inserted on surface", or
assertion failure "A facet is not in conflict with its refinement point"
* The fix was to use Manifold_with_boundary_tag.
r51871 | palliez
update barycentric coordinates - thanks Andreas.
r51869 | lsaboret
Fixed assertion failure "tester(d)" in Triangulation_3::find_conflicts() when running the Delaunay refinement:
* To reproduce the bug:
> poisson_reconstruction data/sphere926.pwn sphere926-poisson.off
(try several times)
* The fix is:
- implement Robust_circumcenter_filtered_traits_3::compute_squared_radius_3_object()
- compare denominator with 1E-13 (instead of 1E-14) to switch to exact arithmetic
r51865 | lsaboret
Fixed infinite loop in Min_sphere_d:
* On Linux/g++, the bug could be reproduced by:
> poisson_reconstruction data/sphere926.pwn sphere926-poisson.off
* The fix was do use Min_sphere_of_spheres_d instead of Min_sphere_d
........
r51901 | palliez | 2009-09-11 09:13:41 +0200 (Fri, 11 Sep 2009) | 1 line
PCA: undo recent change - there was no bug
........
r51902 | lrineau | 2009-09-11 10:44:55 +0200 (Fri, 11 Sep 2009) | 16 lines
Remove mention to piecewise smooth surfaces. The following revision forgot
to remove those two paragraphs:
| ------------------------------------------------------------------------
| r46327 | lrineau | 2008-10-17 14:35:23 +0200 (Fri, 17 Oct 2008) | 3 lines
| Changed paths:
| M /trunk/Surface_mesher/doc_tex/Surface_mesher/main.tex
| M /trunk/Surface_mesher/doc_tex/Surface_mesher_ref/SurfaceMeshFacetsCriteria_3.tex
| M /trunk/Surface_mesher/doc_tex/Surface_mesher_ref/Surface_mesh_complex_2_in_triangulation_3.tex
| M /trunk/Surface_mesher/doc_tex/Surface_mesher_ref/intro.tex
| M /trunk/Surface_mesher/doc_tex/Surface_mesher_ref/main.tex
|
| Update documentation for CGAL-3.4: remove any occurrence of "polyhedral
| [surfaces]" and "piecewise [smooth surfaces]"
|
| ------------------------------------------------------------------------
........
r51931 | stayeb | 2009-09-14 15:22:12 +0200 (Mon, 14 Sep 2009) | 1 line
Fix bbox_3_segment_3 intersection for filtered kernels.
........
r51932 | lrineau | 2009-09-14 15:32:24 +0200 (Mon, 14 Sep 2009) | 3 lines
Fix predicates of <CGAL/AABB_intersection.h>, so that they are correctly
filtered when called from the global function template CGAL::do_intersect
........
r51936 | stayeb | 2009-09-14 19:37:18 +0200 (Mon, 14 Sep 2009) | 2 lines
* Fix ray_bbox & line_bbox do_intersect predicates for filtered kernels
* add a small test for do_intersect(seg/ray/line, bbox) queries
........
r51937 | stayeb | 2009-09-14 19:52:01 +0200 (Mon, 14 Sep 2009) | 1 line
Fix polyhedral oracle issue: switch to exact intersection computation.
........
r51939 | stayeb | 2009-09-15 10:56:32 +0200 (Tue, 15 Sep 2009) | 1 line
Fix minor bug in aabb_do_intersect_test.cpp
........
r51948 | spion | 2009-09-15 19:09:03 +0200 (Tue, 15 Sep 2009) | 5 lines
Rename following the converging convention to prevent future incompatibility :
- Vertex_container -> Vertex_range
- Face_container -> Face_range
- Cell_container -> Cell_range
........
r51953 | stayeb | 2009-09-16 13:17:12 +0200 (Wed, 16 Sep 2009) | 1 line
Fix AABB_tree missing default ctor.
........
r51954 | stayeb | 2009-09-16 14:40:50 +0200 (Wed, 16 Sep 2009) | 6 lines
Merge [r51090]
From trunk/Mesh_3 to branches/CGAL-3.5-branch/Mesh_3.
r51090 | stayeb
* Labels are now ordered according to their natural order (instead of randomly) in medit output.
........
r51957 | stayeb | 2009-09-17 10:20:36 +0200 (Thu, 17 Sep 2009) | 1 line
cleanup example folder.
........
r51960 | stayeb | 2009-09-17 13:10:50 +0200 (Thu, 17 Sep 2009) | 1 line
Remove wrong using namespace statement.
........
r51999 | lsaboret | 2009-09-22 10:27:10 +0200 (Tue, 22 Sep 2009) | 1 line
Removed debugging traces
........
r52026 | lrineau | 2009-09-28 14:18:10 +0200 (Mon, 28 Sep 2009) | 2 lines
Add a minor changelog from CGAL-3.5-beta1.
........
r52075 | lrineau | 2009-09-30 17:24:08 +0200 (Wed, 30 Sep 2009) | 2 lines
Fix the display of the manual testsuite: pdflatex before bibtex and so on.
........
r52079 | lrineau | 2009-10-01 00:21:47 +0200 (Thu, 01 Oct 2009) | 1 line
Fix links to Boost property maps documentation.
........
r52094 | lrineau | 2009-10-01 17:43:44 +0200 (Thu, 01 Oct 2009) | 2 lines
Fix the background of dev manual
........
r52095 | lrineau | 2009-10-01 18:06:36 +0200 (Thu, 01 Oct 2009) | 4 lines
#include <cstdio> is needed, with recent g++ compilers, to find the BUFSIZ
macro.
........
r52096 | lrineau | 2009-10-01 18:07:03 +0200 (Thu, 01 Oct 2009) | 3 lines
Detect images that are linked with background=...
........
r52101 | lrineau | 2009-10-01 18:32:46 +0200 (Thu, 01 Oct 2009) | 2 lines
Fix the bibtex entry of Mesh_3: incorrect title
........
r52102 | lrineau | 2009-10-01 18:34:59 +0200 (Thu, 01 Oct 2009) | 2 lines
Real fix of the bib entry of Mesh_3!
........
r52103 | lrineau | 2009-10-01 18:36:08 +0200 (Thu, 01 Oct 2009) | 1 line
Commit of changes to how_to_cite in the manual tools... (fix the mesh_3 entry)
........
r52104 | lrineau | 2009-10-01 18:41:39 +0200 (Thu, 01 Oct 2009) | 2 lines
Error checking
........
r52105 | lrineau | 2009-10-01 18:49:38 +0200 (Thu, 01 Oct 2009) | 2 lines
Add README files about the generation of the how_to_cite_cgal stuff.
........
r52110 | lrineau | 2009-10-02 10:55:03 +0200 (Fri, 02 Oct 2009) | 2 lines
Fix HTML errors in the manual testsuite results pages.
........
r52111 | lrineau | 2009-10-02 11:11:46 +0200 (Fri, 02 Oct 2009) | 3 lines
Add the link to the HTML validator. The manual testsuite pages validate,
now.
........
r52115 | lrineau | 2009-10-02 16:47:37 +0200 (Fri, 02 Oct 2009) | 2 lines
Minor fix of the output of the manual testsuite.
........
r52141 | lrineau | 2009-10-05 14:28:24 +0200 (Mon, 05 Oct 2009) | 2 lines
Patch from Olivier Devillers: add a check in is_valid.
........
r52142 | lrineau | 2009-10-05 14:33:01 +0200 (Mon, 05 Oct 2009) | 2 lines
CGAL-3.5 published. Let's target 3.5.1, now.
........
r52143 | lrineau | 2009-10-05 14:39:18 +0200 (Mon, 05 Oct 2009) | 2 lines
Fix a link in the bibtex file.
........
r52150 | lrineau | 2009-10-05 19:01:43 +0200 (Mon, 05 Oct 2009) | 3 lines
Update the LaTeX label for "3D Periodic Triangulations" (the previous one
was a copy-paste from TDS_3).
........
r52151 | reichel | 2009-10-05 22:47:27 +0200 (Mon, 05 Oct 2009) | 2 lines
update Debian stuff for 3.5-1
........
r52197 | lrineau | 2009-10-07 21:57:10 +0200 (Wed, 07 Oct 2009) | 2 lines
Boost 1.33.1 *or higher* is needed.
........
r52582 | mcaroli | 2009-10-16 10:19:24 +0200 (Fri, 16 Oct 2009) | 3 lines
bugfix I/O operator
merge of revision 52311 from trunk
........
r52584 | mcaroli | 2009-10-16 10:45:08 +0200 (Fri, 16 Oct 2009) | 3 lines
minor improvements in the reference documentation.
this is a merge from the trunk
........
r52618 | mcaroli | 2009-10-19 17:50:53 +0200 (Mon, 19 Oct 2009) | 2 lines
move the new check in is_valid to the right place (cf. r52141)
........
r52733 | mcaroli | 2009-10-26 10:15:53 +0100 (Mon, 26 Oct 2009) | 11 lines
Merge revisions r52701 and r52704 from trunk:
| ------------------------------------------------------------------------
| r52701 | mcaroli | 2009-10-23 11:15:57 +0200 (Fri, 23 Oct 2009) | 2 lines
|
| add the domain to the basic example
| ------------------------------------------------------------------------
| r52704 | mcaroli | 2009-10-23 11:56:56 +0200 (Fri, 23 Oct 2009) | 2 lines
|
| add a phrase that precises that input points have to lie inside the cube
| ------------------------------------------------------------------------
........
r52843 | lrineau | 2009-11-02 14:28:09 +0100 (Mon, 02 Nov 2009) | 2 lines
g++-4.4 is supported as well.
........
r53002 | mcaroli | 2009-11-12 18:05:01 +0100 (Thu, 12 Nov 2009) | 3 lines
improved doc of periodic_point(v) and periodic_point(v,i)
this is a merge from the trunk, revision 53001
........
r53053 | stayeb | 2009-11-16 13:56:49 +0100 (Mon, 16 Nov 2009) | 5 lines
Merge r53048 from trunk
r53048 | stayeb
Fix syntax error.
........
r53152 | stayeb | 2009-11-24 13:43:03 +0100 (Tue, 24 Nov 2009) | 1 line
Fix bug (typo) in polyhedral domain initialization.
........
r53292 | lrineau | 2009-12-04 16:00:00 +0100 (Fri, 04 Dec 2009) | 1 line
commit revision 53291 from trunk to branch 3.5
........
r53293 | lrineau | 2009-12-04 16:01:30 +0100 (Fri, 04 Dec 2009) | 1 line
commit revision 53289\ from trunk to branch 3.5
........
r53294 | lrineau | 2009-12-04 16:02:06 +0100 (Fri, 04 Dec 2009) | 1 line
commit revision 53288 from trunk to branch 3.5
........
- Deprecate the "hierarchy" implementation detail in the package API.
- Introduce a Location_policy parameter instead.
- Add a "Complexity and Performance" section in the user manual.
- Add a benchmark program (dont_submit'ed for now)
Note : merge done by hand as svnmerge.py got too easily confused...
r19107 | afabri | 2003-10-17 10:49:19 +0200 (Ven 17 oct 2003) | 2 lignes
Added const& for gaining performance
which was justified at the time by the fact that on VC++, handles encapsulated iterators.
See Andreas' e-mail:
> I just had a look at the code. The problem is that it calls
> incident_cells, which is implemented recursively, and for a
> vertex with many incident cells, as in your case the infinite
> vertex, the stack is full.
>
> We have to put it on our todo list.
I did it for 3D only because the degenerate 2D case should be
handled by the circulator anyway.
I did not add the test which explodes the call stack (in case we plug
the recursive version): too slow for a testsuite. But incident_...
methods are used everywhere in the code anyway.
- all the code is factored around visit_incident_cells.
- incident_edges has been added
- finite versions of all incident_*(vertex) have been added
thanks to a Filter template parameter
det2x2_by_formula
det3x3_by_formula
det4x4_by_formula
det5x5_by_formula
det6x6_by_formula
to:
determinant
How cute... a name independent of the dimension, and even readable !
sign_of_determinant2x2
sign_of_determinant3x3
sign_of_determinant4x4
sign_of_determinant5x5
sign_of_determinant6x6
to:
sign_of_determinant
So that we have less dimension-dependent namings, at least internally...
* CGAL_error to CGAL_error_msg
* introduced a macro CGAL_error()
* added some words about CGAL_error to the developers manual
* renamed most of assert(x) into CGAL_assertion(x)
* renamed exit(x) with x != 0 , CGAL_assertion(false) and assert(false) into CGAL_error
* CORE left untouched, OpenNL changed
- Add an undocumented circumcenter() template member function
in Triangulation_ds_cell_base_3 so that we can continue to use it as default
template parameter of the TDS, and still have .dual() working when only the
VertexBase type is changed (e.g. for the hierarchy).
Sébastien Loriot
Laurent Rineau
Andreas Fabri
Monique Teillaud
Sylvain Pion
Stéphane Tayeb
Olivier Devillers
Manuel Caroli
Camille Wormser
Fernando Cacciola
Andreas Meyer
Christophe Delage