Usage of mmg3d does not work like in doc

General Discussion
1

Hello,
I tried to use mmg3d to adapt some 3d mesh. However, it seems that the examples in the documentation pdf are not up to date. I use version 4.7 of FreeFEM.

I tried to use the command

Th3 = mmg3d(Th,options=opt);
with opt being an int[int] of size 6 as shown in the documentation.

However FreeFem gives an error that there is no option “options”. I see “opt” as an option, but it does not work. Can you please give an example of usage of mmg3d inside FreeFEM? Thank you.

Best regards,
Beniamin

2

You can do the following:

$ git clone -b develop https://github.com/FreeFem/FreeFem-sources
$ cd FreeFem-sources
$ git grep "load \"mmg\""
examples/3d/Laplace-Adapt-aniso-3d.edp:load "mmg"
examples/3d/Sphere-Isocahedron.edp:load "mmg"
examples/3dSurf/Pinocchio.edp:load "mmg"
examples/hpddm/laplace-adapt-3d-PETSc.edp:load "mmg"
examples/hpddm/transfer.edp:load "mmg"

You can look at any of these examples to see how to use Mmg.

1 Like
3

Thank you for the answer. I was able to adapt one of the examples to what I am doing.

4

Hello,

In the souce code it appears that :

basicAC_F0::name_and_type mmg3d_Op::name_param = {
{“metric”, &typeid(KN< double > *)}, // 0
{“displacement”, &typeid(E_Array)}, // 1
{“displVect”, &typeid(KN_< double >)}, // 2
{“opt”, &typeid(string *)}, // 3
{“Mb”, &typeid(long)} // 4
};

  • Memory usage is controled with “Mb”
  • Options are defined using “opt” string

Does anybody know the options in the string ?

Best regards.

5

Hello,

In the souce code it appears that :

basicAC_F0::name_and_type mmg3d_Op::name_param = {
{“metric”, &typeid(KN< double > *)}, // 0
{“displacement”, &typeid(E_Array)}, // 1
{“displVect”, &typeid(KN_< double >)}, // 2
{“opt”, &typeid(string *)}, // 3
{“Mb”, &typeid(long)} // 4
};

  • Memory usage is controled with “Mb”
  • Options are defined using “opt” string

Does anybody know the options in the string ?

As Prof. Pierre suggested, you should use the new “mmg” module (~v5.5.2) instead of the old “mmg3d-v4.0” module (~v4.0). These parameters you mentioned exist in mmg3d-v4.0 (which is an old version Mmg in 3d). If you just want to understand what these old parameters mean, you can refer to P168 in FreeFEM tutorial https://doc.freefem.org/pdf/FreeFEM-documentation.pdf.

6

Hi,
I use the latest version of freefem (4.9), it includes mmg3d 4.0. Do you know where to find dll file of mmg 5.5.2 (cause the mmg website only offer exe) ?

7

I use the latest version of freefem (4.9), it includes mmg3d 4.0. Do you know where to find dll file of mmg 5.5.2 (cause the mmg website only offer exe) ?

In FreeFem-sources/plugin/seq/mmg.cpp

8

That’s the source code… not sure to understand.

9

mmg.so is the dll file, which is generated by mmg.cpp. If you want to refer to all parameters added, you also can see mmg.cpp.

10

Sorry, i made a confusion, i believed that mmg.dll was for 2d and mmg3d for 3d… now it’s fine, mmg.dll contains both 2d and 3d !
Thanks.

11

Yes, the new version Mmg include mmgs and mmg3d

12

I try to use mmg3d for mesh adaption using mshmet first (to get the metric from a scalar solution) but i’m not really sure.
Did you ever used mmg for 3d mesh refinement ?

13

I try to use mmg3d for mesh adaption using mshmet first (to get the metric from a scalar solution) but i’m not really sure.

Right, using “Mshmet + Mmg” if you use the default metric instead of the metric constructed by yourself.

Did you ever used mmg for 3d mesh refinement ?

Yes, I used them to do mesh adaptation

14

I have an error during mmg 3d for a classic laplacian equation on an anisotropic mesh

h = mshmet(Th, u, normalization=true); // u is the initial solution
Th = mmg3d(Th, metric=h, mem=1000,hmin=1e-3, hmax=1e-1);

– PHASE 2 : ISOTROPIC MESHING
Assertion failed: MG_GET(pxt1->ori,voyq ), file C:/builds/workspace/deployEXE/3rdparty/ff-petsc/petsc-3.15.0/fr/externalpackages/git.mmg/src/mmg3d/colver_3d.c, line 1131
Unexpected error: *** potential lack of memory.

How to use mm3g for anisotropic meshes ?

15

How to use mm3g for anisotropic meshes ?

As Prof. Pierre (prj) suggested before, you can refer to the following examples to study how to use it.

examples/3d/Laplace-Adapt-aniso-3d.edp:load “mmg”
examples/3d/Sphere-Isocahedron.edp:load “mmg”
examples/3dSurf/Pinocchio.edp:load “mmg”
examples/hpddm/laplace-adapt-3d-PETSc.edp:load “mmg”
examples/hpddm/transfer.edp:load “mmg”

16

Yes that very good examples, i do the same.
When i activate aniso in mshmet, the metric is very strange, but with aniso=0 the metric looks better.
I want a final anisotropic mesh, so i guess i must set aniso=true ? (as you see, my initial mesh is highly anisotropic, it contains thin layers)

Moreover mmg3d doesn’t work whatever aniso.

Here my very basic script with a .sol file

Can you have a look ? Thank you for your help.

adapt.zip (192.9 KB)

17

Here the modified edp script with a finer mesh. The script crash when performing mmg3d.

adapt3d.zip (184.2 KB)

18

Yes that very good examples, i do the same.
When i activate aniso in mshmet, the metric is very strange, but with aniso=0 the metric looks better.
I want a final anisotropic mesh, so i guess i must set aniso=true ? (as you see, my initial mesh is highly anisotropic, it contains thin layers)

Moreover mmg3d doesn’t work whatever aniso.

Here my very basic script with a .sol file

Can you have a look ? Thank you for your help.

I captured two errors:

  1. hmet in your temp.edp is in P1, but it should be in [P1, P1, P1, P1, P1, P1]
  2. I’m not sure that loading your temp.mesh is right (there seems to be a problem in your temp.mesh or loading ‘temp.mesh’). For the temp.mesh and its loading, I don’t know.
19

Hello,

Thanks yo for your answer.

This two blocks are different ?

fespace Mh(Th,[P1, P1, P1, P1, P1, P1]);
Mh [m11,m21,m22,m31,m32,m33];
real[int] met = mshmet(Th, u, hmin=1e-3,hmax=0.2, err=errm, aniso=1);
m11 = met;

fespace Mh(Th,P1);
Mh m11;
real[int] met = mshmet(Th, u, hmin=1e-3,hmax=0.2, err=errm, aniso=1);
m11 = met;

the scalar field m11 looks very strange for a metric…

image
image732×787 43.3 KB

Concerning the mesh :
readmesh3 with the medit file temp.mesh indicates one warning

— Warning manifold obj nb:432 adj 446 of dim =2

The mesh comes from a brep file that i mesh with gmsh and export it as medit file. You can find file the brep file here.
I tried to use gmsh mesh file (gmshload3, no warning when loading), mmg3d doesn’t crash but it’s still running (more than 15min, then i stopped) with some worrying warnings…

Warning: MMG5_elementWeight: at least 1 negative or null density.
Warning: MMG5_movbdyregpt_ani: unable to compute optimal position for at least 1 point.

Sorry for disturbing you.

Best regards.

20

This two blocks are different ?

The second code block is wrong, and please note that met obtained by Mshmet describe the metric at each vertice in the mesh (It’s a flatten vector). Checking if m11[].n equals met.n before doing m11[] = met;, you can also find the wrong from this point.

Sorry, I don’t know about loading and exporting of Gmsh grids and the gmshload3 module