On Sun, 12 Jan 2014 13:08:27 +0000 "Alain M. Clo" <[email protected]> wrote:
> Dear Jan > How does Dolphin take advantage of threads (OpenMP) ? > I would like to parallelize the solver par. > Is this implemented and what is the interface ? > I checked by adding : > parameters["num_threads"] = "16" This ensures that threaded OpemMPAssembler is used for assembly (if your DOLFIN is compiled with OpenMP and coloring library (I think boost is used)). Problem is the threaded solver. Supported one is PaStiX, but it is difficult to compile and use. I think that PETSc is also gradually getting some thread support but I don't think it is supported from the side of DOLFIN. As a result you can see an advantage of using OpenMP for assembly-extensive applications. On the other hand, MPI parallelism is supported to the large extent. If you have DOLFIN compiled with MPI and parallel algebra backend (PETSc, Trilinos), you can just try to run demos parallely $ mpirun -n 16 python demo_stokes-iterative.py to run on 16 proccesors (for example on 2 eight-core machines). Just note that you will not see an adequate speed-up on such small demo problems. Refine used mesh to take advantage of this. Jan > to the python script > but this was not successful > Thanks > Alain > ________________________________________ > From: Jan Blechta [[email protected]] > Sent: Sunday, January 12, 2014 1:55 PM > To: Alain M. Clo > Subject: Re: [FEniCS-support] Fenics demo on MultiPhysics and Fenics > on RedHat6 > > On Fri, 10 Jan 2014 22:37:08 +0000 > "Alain M. Clo" <[email protected]> wrote: > > > Dear Jan, > > I have another question regarding FEniCS and FreeFem++. > > I am assuming that you know FreeFem++. > > This is wrong assumption. > > Jan > > > A part from the added interface of python in FEniCS, > > Where do you see them differ ? strength/weaknesses ... > > (It seems to me that FEniCS also has a larger adoption through the > > World) Thanks > > Alain > > ________________________________________ > > From: Alain M. Clo > > Sent: Friday, January 10, 2014 5:06 PM > > To: Jan Blechta > > Subject: RE: [FEniCS-support] Fenics demo on MultiPhysics and Fenics > > on RedHat6 > > > > OK :-) > > I will do a coupled case then. > > Thanks > > Alain > > ________________________________________ > > From: Jan Blechta [[email protected]] > > Sent: Friday, January 10, 2014 4:52 PM > > To: Alain M. Clo > > Cc: [email protected] > > Subject: Re: [FEniCS-support] Fenics demo on MultiPhysics and Fenics > > on RedHat6 > > > > On Fri, 10 Jan 2014 13:40:59 +0000 > > "Alain M. Clo" <[email protected]> wrote: > > > > > Hello Jan and Johannes, > > > > > > Thanks for the tips on the demo. > > > Almost all run perfectly well. > > > The documentation is very nice to read, and the number of demos is > > > impressive, all is high quality. > > > > > > I have one question : > > > Which demo case is a MultiPhysic case ? > > > > There is no proper multiphysics demo because of > > http://fenicsproject.org/pipermail/fenics-support/2014-January/000268.html > > > > Jan > > > > > I haven't seen more than one physic case. > > > Thanks to point the case to me, and I will take a second look. > > > Kind Regards > > > > > > Alain > > > > > > ________________________________________ > > > From: Jan Blechta [[email protected]] > > > Sent: Monday, January 06, 2014 6:24 PM > > > To: Alain M. Clo > > > Cc: [email protected] > > > Subject: Re: [FEniCS-support] Fenics demo on MultiPhysics and > > > Fenics on RedHat6 > > > > > > On Mon, 6 Jan 2014 13:47:05 +0000 > > > "Alain M. Clo" <[email protected]> wrote: > > > > > > > Hello Jan and Johannes, > > > > Thanks for the response. > > > > Are the demo mentioned below available and plug and play type > > > > with just scripts available ? Or does it assume some > > > > developments ... > > > > > > Having FEniCS correctly installed, initialize your shell > > > environment for FEniCS (depends on a type of installation) and > > > type something like > > > > > > $ > > > cd /path/to/FEniCS/share/dolfin/demo/documented/navier-stokes/python > > > $ python demo_navier-stokes.py > > > > > > and enjoy! > > > > > > Some demos are extensively documented > > > http://fenicsproject.org/documentation/dolfin/dev/python/demo/index.html > > > http://fenicsproject.org/documentation/dolfin/1.2.0/python/demo/index.html > > > > > > Jan > > > > > > > (meshing, boundary conditions ...) My demo is in 10 days, so > > > > there is not much time left for me then. Alain > > > > > > > > ________________________________________ > > > > From: Jan Blechta [[email protected]] > > > > Sent: Monday, January 06, 2014 4:08 PM > > > > To: Alain M. Clo > > > > Cc: Johannes Ring; [email protected] > > > > Subject: Re: [FEniCS-support] Fenics demo on MultiPhysics and > > > > Fenics on RedHat6 > > > > > > > > On Mon, 6 Jan 2014 11:24:35 +0000 > > > > "Alain M. Clo" <[email protected]> wrote: > > > > > > > > > Hello Johannes > > > > > Thanks for the feedback > > > > > Do you mean buy support to install it ? > > > > > Or should it install without problem with the dev version ? > > > > > Currently it is running on Ubuntu > > > > > And I can do the demo on that platform, > > > > > Later we will need it on RH6. > > > > > > > > > > Any idea on a multiphysics case ? > > > > > > > > There are demos of coupled problems (like Stokes, Navier-Stokes, > > > > mixed Poisson, biharmonic, Cahn-Hilliard), solved both > > > > iteratively and fully-coupled, in a directory tree. > > > > > > > > Proper multiphysics, i.e. mixing unknowns on distinct meshes, is > > > > not supported yet, but it is in the focus of developers. > > > > > > > > Usual workaround is to define all the fields of a mixed problem > > > > on the whole domain and restrict them by an application of > > > > DirichletBC by arbitrary value at spare regions. This can be a > > > > little bit tricky sometimes. There are also more advanced > > > > approaches like telling to PETSc what is null space of your > > > > operator (like in singular-poisson demo). > > > > > > > > On the other hand if you decouple your problem into a sequence > > > > of single-mesh problems, then you can solve coupled problem > > > > iteratively without any workaround. > > > > > > > > Jan > > > > > > > > > > > > > > Regards > > > > > Alain > > > > > > > > > > Le 6 janv. 2014 à 14:10, "Johannes Ring" <[email protected]> a > > > > > écrit : > > > > > > > > > > > On Thu, Jan 2, 2014 at 5:44 PM, Alain M. Clo > > > > > > <[email protected]> wrote: > > > > > >> Hello Fenics Support, > > > > > >> > > > > > >> I have 2 requests : > > > > > >> -1 Multiphysics demo > > > > > >> -2 FEniCS on Redhat 6 > > > > > >> > > > > > >> 1-I am from Kaust and planning a short tutorial on Fenics > > > > > >> I will make a demo using the Lapace/Poisson example > > > > > >> and am looking also introducing an example of > > > > > >> MultiPhysics. Do you something ready that I could demo ? > > > > > >> Or help build me a simple case > > > > > >> > > > > > >> 2-I had some difficulties installing on RH6 > > > > > >> Are there some notes on practicing on RH6 with FEniCS ? > > > > > > > > > > > > No, but Dorsal has support for installing FEniCS on this > > > > > > platform: > > > > > > > > > > > > http://fenicsproject.org/download/installation_using_dorsal.html > > > > > > > > > > > > Currently, it is recommended to build the development > > > > > > version of FEniCS. You can do that by setting > > > > > > STABLE_BUILD=false in dorsal.cfg. > > > > > > > > > > > > Johannes > > > > > > > > > > ________________________________ > > > > > > > > > > This message and its contents including attachments are > > > > > intended solely for the original recipient. If you are not > > > > > the intended recipient or have received this message in > > > > > error, please notify me immediately and delete this message > > > > > from your computer system. Any unauthorized use or > > > > > distribution is prohibited. Please consider the environment > > > > > before printing this email. > > > > > _______________________________________________ > > > > > fenics-support mailing list [email protected] > > > > > http://fenicsproject.org/mailman/listinfo/fenics-support > > > > > > > > > > > > ________________________________ > > > > > > > > This message and its contents including attachments are intended > > > > solely for the original recipient. If you are not the intended > > > > recipient or have received this message in error, please notify > > > > me immediately and delete this message from your computer > > > > system. Any unauthorized use or distribution is prohibited. > > > > Please consider the environment before printing this email. > > > > _______________________________________________ fenics-support > > > > mailing list [email protected] > > > > http://fenicsproject.org/mailman/listinfo/fenics-support > > > > > > > > > ________________________________ > > > > > > This message and its contents including attachments are intended > > > solely for the original recipient. If you are not the intended > > > recipient or have received this message in error, please notify me > > > immediately and delete this message from your computer system. Any > > > unauthorized use or distribution is prohibited. Please consider > > > the environment before printing this email. > > > _______________________________________________ fenics-support > > > mailing list [email protected] > > > http://fenicsproject.org/mailman/listinfo/fenics-support > > > > > > ________________________________ > > > > This message and its contents including attachments are intended > > solely for the original recipient. If you are not the intended > > recipient or have received this message in error, please notify me > > immediately and delete this message from your computer system. Any > > unauthorized use or distribution is prohibited. Please consider the > > environment before printing this email. > > > ________________________________ > > This message and its contents including attachments are intended > solely for the original recipient. If you are not the intended > recipient or have received this message in error, please notify me > immediately and delete this message from your computer system. Any > unauthorized use or distribution is prohibited. Please consider the > environment before printing this email. _______________________________________________ fenics-support mailing list [email protected] http://fenicsproject.org/mailman/listinfo/fenics-support
