[Gmsh] How do I set the random number seed with a "Mesh.Seed" option for deterministic mesh generation?

[Thomas Kiel]

Dear Max, dear Christophe,

thank you for the fast answers. As it turned out I made a stupid mistake 
and forgot to replace a symbolic link still pointing the gmsh 3.0.6 
executable, while it actually should have pointed to the fresh 
downloaded binary of 4.0.4. This now gives me the same meshes after 
running gmsh in batch mode several times. However, adding your 
Mesh.RandomFactor entries did not reproduce the surface mesh warning on 
my machine.

Thanks a lot,

Thomas

Am 2018-10-27 09:56, schrieb Christophe Geuzaine:
>> On 27 Oct 2018, at 00:16, Max Orok <morok at mevex.com> wrote:
>> 
>> Hi Thomas,
>> 
>> I first tried to set the random factors using:
>> Mesh.RandomFactor = 1e-9;
>> Mesh.RandomFactor3D = 1e-9;
>> 
>> I think the bug is elsewhere however...
>> With your file, I get a warning the first time and none the subsequent 
>> times.
>> I think it is due to internal state with the mesher remembering the 
>> choice of MeshAdapt?
> 
> Yes : when meshing fails with Delaunay or Frontal we currently force
> MeshAdapt on that surface. Currently this choice is persistent (in a
> Gmsh run) - I will change this so that the initial state is restored.
> 
> 
>> Switching from Delauney to MeshAdapt halfway might cause different 
>> point spacing etc.
>> 
>> Here are some pictures showing the two consecutive mesh logs
>> 
>> 1st time:
>> <image.png>
>> 
>> 2nd time (same window):
>> <image.png>
>> 
>> On closing Gmsh and trying again, this repeats.
>> 
>> Instead, to fix your problem, you can set the meshing algo to 
>> MeshAdapt ahead of time by changing the line here:
>> Mesh.Algorithm = 1; // changed to MeshAdapt
>> 
>> Three repeated trials gave identical output files.
>> 
>> p.s. just in case, all of the current / modified options can be looked 
>> through and searched using Ctrl-Shift-H or Help -> Current Options and 
>> Workspace
>> 
>> Sincerely,
>> Max
>> 
>> On Fri, Oct 26, 2018 at 12:44 PM, Thomas Kiel 
>> <Thomas.Kiel at physik.hu-berlin.de> wrote:
>> Dear Gmshers,
>> 
>> as I already pointed out in an earlier email (s. 
>> http://onelab.info/pipermail/gmsh/2018/012045.html), I had problems 
>> with reproducing the exact same mesh twice. Since this email was 
>> unfortunately html encoded I think it was overlooked. Yet I like to 
>> apologize to ask again, this time in text formatted email, how to set 
>> the random number generator seed with an outside option?
>> 
>> I know, that a grep to "srand" within the gmsh git repository gives 
>> several times hard-coded initialized seeds, e.g.
>> 
>> Geo/SOrientedBoundingBox.cpp:326:  srand((unsigned)time(0));
>> Mesh/Levy3D.cpp:624:  srand(time(NULL));
>> Mesh/Levy3D.cpp:1359:    srand(time(NULL));
>> Mesh/meshGFaceRecombine.cpp:245:  srand(time(NULL));
>> Mesh/meshGFaceLloyd.cpp:394:  srand(time(NULL));
>> 
>> yet I am not sure whether these influence my mesh generation. A 
>> minimalistic test file which causes different meshes at consecutive 
>> runs in my case, is
>> 
>> 
>> //+
>> SetFactory("OpenCASCADE");
>> 
>> Mesh.CharacteristicLengthMin = 0.001;
>> Mesh.CharacteristicLengthMax = 2.0;
>> 
>> Mesh.CharacteristicLengthExtendFromBoundary = 1;
>> 
>> Mesh.Algorithm = 5;
>> Mesh.Smoothing = 3;
>> Mesh.Algorithm3D = 1;
>> 
>> Box(1) = {0, -0.5, -0.5, 1, 1, 1};
>> Box(2) = {-1, -0.5, -0.5, 1, 1, 1};
>> Sphere(3) = {0.5, 0.0, 0.0, 0.3, -Pi/2, Pi/2, 2*Pi};
>> 
>> Coherence;
>> 
>> surfs_box1() = Unique(Abs(Boundary{ Volume{4}; } ));
>> lines_box1() = Unique(Abs(Boundary{ Surface{surfs_box1()}; } ));
>> pts_box1() = Unique(Abs(Boundary{ Line{lines_box1()}; } ));
>> 
>> surfs_sphere() = Unique(Abs(Boundary{ Volume{3}; } ));
>> lines_sphere() = Unique(Abs(Boundary{ Surface{surfs_sphere()}; } ));
>> pts_sphere() = Unique(Abs(Boundary{ Line{lines_sphere()}; } ));
>> 
>> surfs_box2() = Unique(Abs(Boundary{ Volume{2}; } ));
>> lines_box2() = Unique(Abs(Boundary{ Surface{surfs_box2()}; } ));
>> pts_box2() = Unique(Abs(Boundary{ Line{lines_box2()}; } ));
>> 
>> // sorted from largest to smallest:
>> Characteristic Length{pts_box2()} = 0.5;
>> Characteristic Length{pts_sphere()} = 0.2;
>> Characteristic Length{pts_box1()} = 0.1;
>> 
>> 
>> 
>> I tested this on a machine running Ubuntu 16.04 and installed the 
>> out-of-the-box gmsh 4.0.4 version from the webpage. If any further 
>> detail on the running machine is required please tell me.
>> I hope you can help me out.
>> 
>> Thanks a lot in advance and best regards,
>> 
>> Thomas Kiel
>> 
>> --
>> Thomas Kiel
>> Theoretische Optik und Photonik
>> 
>> Humboldt-Universität zu Berlin, Institut für Physik
>> Newtonstraße 15, 12489 Berlin
>> 
>> Tel.:   +49 30 2093 7999
>> E-Mail: thomas.kiel at physik.hu-berlin.de
>> Web:    https://top.physik.hu-berlin.de/people/thomas-kiel
>> 
>> _______________________________________________
>> gmsh mailing list
>> gmsh at onelab.info
>> http://onelab.info/mailman/listinfo/gmsh
>> 
>> 
>> 
>> --
>> Max Orok
>> Contractor
>> www.mevex.com
>> 
>> 
>> _______________________________________________
>> gmsh mailing list
>> gmsh at onelab.info
>> http://onelab.info/mailman/listinfo/gmsh
> 
> —
> Prof. Christophe Geuzaine
> University of Liege, Electrical Engineering and Computer Science
> http://www.montefiore.ulg.ac.be/~geuzaine
> 
> Free software: http://gmsh.info | http://getdp.info | 
> http://onelab.info

-- 
Thomas Kiel
Theoretische Optik und Photonik

Humboldt-Universität zu Berlin, Institut für Physik
Newtonstraße 15, 12489 Berlin

Tel.:   +49 30 2093 7999
E-Mail: thomas.kiel at physik.hu-berlin.de
Web:    https://top.physik.hu-berlin.de/people/thomas-kiel