Christophe GeuzaineThe Second Gmsh Workshop took place in Moressée (Belgium) on May 23-24 2013.
La Ferme de la Bourgade, Moressée, 10, 5377 Heure (Belgique) (Map)
Programme
Wednesday May 22
- 18:00 Arrival for those who wish to stay the night
Thursday May 23
- 9:00 - Welcome - Coffee
- 9:30 - 10:00 Introduction - What's new in Gmsh since the first workshop? (C. Geuzaine, ULg)
- 10:00 - 12:45 Presentations about different projects in the Gmsh framework
- Mesh generation in Gmsh: a roadmap (J.-F. Remacle, UCL)
- High-order bounds (A. Johnen, ULg) and High-order meshing (T. Toulorge, UCL)
- Fast assembly of non-Lagrange elements (N. Marsic, ULg)
- Homology and cohomology solver in Gmsh and its applications (M. Pellikka, TUT)
- Derivation of the Gmsh DofManager for general weak coupled system resolutions, application to non local damage (G. Becker, MIT)
- ONELAB, metamodels and Python (F. Henrotte, UCL)
- Gmsh and ONELAB on Android and iOS (M. Graulich, ULg/ISET)
- 12:45 - 14:00 Lunch
- 14:00 - 16:00 Mini-workshops (1)
- Interfacing solvers with ONELAB (existing and home-made, C++ and Python)
- Advanced mesh generation techniques (high-order meshes, tubular structures for biomedical applications, quad/hex meshes and boundary layers)
- 16:00 - 18:00 Social activity
- 18:00 - 19:00 Talk around a pre-dinner drink
- Invited lecture: Gmsh @ NECS (Shahrokh Ghavamian, NECS)
- 19:30 Dinner
Friday May 24
- 9:30 - 10:30 Breakfast talks
- Invited lecture: Code_Aster (Aimery Assire, EDF & ALNEOS)
- Invited lecture: Gmsh as a toolbox for Summit (G. Becker, MIT)
- 10:30 - 10:45 Coffee
- 10:45 - 12:45 Mini-workshops (2)
- Interfacing solvers with ONELAB (existing and home-made, C++ and Python)
- Advanced mesh generation techniques (high-order meshes, tubular structures for biomedical applications, quad/hex meshes and boundary layers)
- 12:45 - 14:00 Lunch
- 14:00 - 16:00 Roadmap for Gmsh:
- Roundtable discussion: What is missing? What should be improved? Who does what? ...
Notes from the roundtable
API:
- we need a stable API; this stable API could defined as
- what's tested by CTest (design by contract; test for invalid input; never segfault)
- what's documented (non-documented or not marked as stable == unstable, not part of API)
- need more self-discipline (coding rules; automate using a two-stage commit process?)
- start by creating testing/ directory with CTests with current database of tests; make it available in SVN; try gcov (Jon's experience is that is takes to long)
- we also need new APIs
- allow to externalize algorithms (meshing, reparam, post-pro; with clear in/out on existing/nicer data classes)
- for "vectorization" (lists of stuff; expose numpy objects? not sure...)
- fix warnings @ compilation for Python bindings
- adding one .h file in the CMakeLists.txt makes cmake recompile everything? Check this (Vincent)
Geometry:
- interface SGEOM @ Salome
- extrusion of splines is bugggggggy as well as extrusion of simple stuff (?)
- re-write/check the parametrizations for extrusion: HO is just impossible with the current ones
- enhance the multi-model switching + clearing
Mesh:
- rendering efficiently of high order meshes (test at corners)
- non equidistant ho elements (gauss lobato)
- document high order stuff; document HO element types (not complete in .texi)
- different orders on different regions (mostly ok)
- extrude meshes from brep surfaces
- boundary layers
- need double BL (both sides)
- not all quality measures are not implemented for all element types
Solver:
- run remote solvers in nohup mode (check if everything still runs)
Post-processing:
- need to restore dynamic plugins (meshing algos in plugins?)
- more than one action in the plugins
- visualization : decimation like in paraview; one single vertex array for one model : would make mesh transparency easy
- multiple time steps should be loaded in an intelligent manner : choose a certain number of time steps, show things while loading the rest of the files
- output a part of the screen (save a viewport)
- zoom in 2D graphs
- saturate colors in a separate color (see Shahrokh's wishlist)
- better manipulation for pre-defined rotations on the screen: +-90 in perp direction
- select what to X-Y plot in onelab is clumsy
Group photo
Local Organizing Committee
- C. Geuzaine (ULg),
- A. Modave (ULg),
- I. Niyonzima (ULg),
- A. Johnen (ULg),
- N. Marsic (ULg)
Scientific Committee
- C. Geuzaine (ULg),
- J.-F. Remacle (UCL),
- R. Sabariego (ULg),
- E. Marchandise (UCL),
- E. Bechet (ULg),
- V. Legat (UCL)