-
task
None
0.00
0%
0
0
I tested the Scilab solvers using demo_tripod_alt.
Here are the results:
The getfem cg solver: 3 sec
The getfem superlu: 4 sec
The Scilab default sparse solver: 261 sec
The Scilab umfpack solver: 1.5 sec
So, this is a really efficient solver. It should be added to getfem.
YC
Unlocked
None
6775
None
ycollet
1254873600
1254873600
Add the umfpack sparse linear solver in getfem
1254907842
Public
1 - Later
1581
Open
-
task
None
0.00
0%
0
0
If we loaded a gmsh mesh file, add the possibility to get set of nodes belonging to a physical (as defined in gmsh)
For example (piece of code extracted from demo_tripod_alt:
* Before:
P = gf_mesh_get(m,'pts');
pidtop = find(abs(P(2,:)-13)<1e-6); // find those on top of the object
pidbot = find(abs(P(2,:)+10)<1e-6); // find those on the bottom
* With physical:
P = gf_mesh_get(m,'pts');
idtop = 101; // the ID of the physical which defines the top of the object
idbot = 102; // the ID of the physical which defines the bottom of the object
pidtop = gf_mesh_get(m,'physical',idtop);
pidbot = gf_mesh_get(m,'physical',idbot);
Unlocked
None
6776
None
ycollet
1254873600
1254873600
If we loaded a gmsh mesh file, add the possibility to get set of nodes belonging to a physical
1254908253
Public
1 - Later
1581
Open
-
task
None
0.00
0%
0
0
The convect method (in src/getfem/getfem_convect.h)
- Should be extended to non Lagrange fem with an additional L2 projection (convecting the Gauss points)
- Should take into account a right hand side by integration on the caracteristic lines.
Unlocked
None
6817
None
renard
1257033600
1351728000
Improvment of the convect method
1257066138
Public
3 - Low
1581
Open
1279635368
Should_be_Finished_on
1257033600
1604188800
renard
1279635407
Should_be_Finished_on
1604188800
1351728000
renard
Priority
5 - Normal
3 - Low
renard
-
task
renard
0.00
80%
0
0
An additional option of the solve should the possibility to solve the model only with respect to a subset of variables.
the consequence is the assembly of sub-tangent matrix.
This should allow to solve a complex coupled problem with a "Gauss-Seidel" approach (or SSOR).
Unlocked
Done
7120
None
renard
1279584000
1500508800
Partial solve of a model
1279635346
Public
5 - Normal
1581
Open
1428158726
Percent_Complete
0%
80%
renard
Should_be_Finished_on
1342742400
1500508800
renard
Original_Submission
The possiblity to make a partial solve has been added by the possibility to disable some variables in the model and solve the model with respect of the remaining variables.
The remaining thing to do concerns the linear bricks coming from the generic assembly. Since the output is a global matrix, it has to be recomputed each time a variable is disabled/enabled again. In order to avoid that, a mechanism allowing to make only one global assembly of linear bricks which dispacth the global matrix into small components matrix (similarly to standard bricks) would be a solution.
renard
Status
None
In Progress
renard
1439553217
Status
In Progress
Done
renard
Assigned_to
100
4654
renard
-
task
None
0.00
20%
0
0
Unlocked
None
7121
None
renard
1279584000
1451520000
Parallelization of Getfem
1279635470
Public
5 - Normal
1581
Open
1279635884
Original_Submission
The Parallelization of Getfem is still incomplete. The assembly procedures of the bricks are parallelized and it is possible to call MUMPS to solve the linear system. To do list:
- Facilitate the install of the parallel version, tests and documentation.
- parallelization of the range Basis algorithm.
- parallelization of the numbering of dofs
- Finish the interface for Schwarz additive domain
decomposition solver.
- See FETI domain decomposition
renard
Priority
1 - Later
5 - Normal
renard
1279635937
Should_be_Finished_on
-14462
1451520000
renard
-
task
logari81
0.00
0%
1471446010
0
The currently available plasticity brick is limited to infinite deformations. However, most problems involving plastic deformation, also include large deformations. A corresponding brick would be very useful.
Are there any recommended literature sources to look at, before starting a new implementation or adapting the currently available brick for plasticity?
Unlocked
Done
7765
None
logari81
1391558400
1423094400
Implement finite deformation plasticity
1391621045
Public
5 - Normal
1581
Closed
1397595939
Original_Submission
some very thorough work on closest point projection method:
https://upcommons.upc.edu/e-prints/bitstream/2117/8253/1/armero_on-the-formulation_2002.pdf
https://upcommons.upc.edu/e-prints/bitstream/2117/8254/1/perez_on-the-formulation_2002.pdf
logari81
1471446010
Status
None
Done
logari81
Open/Closed
Open
Closed
logari81
-
task
logari81
0.00
0%
0
0
Shell elements provide an efficient modelling method for thin solid structures.
Are there any good references to follow for the implementation of triangle and quadrilateral shell elements in getfem?
Unlocked
None
7768
None
logari81
1391990400
1423526400
Implementation of structural shell elements
1392063777
Public
5 - Normal
1581
Open
1392063797
Assigned_to
100
14978
logari81