[aspect-devel] ASPECT on Mac

Menno Fraters menno.fraters at outlook.com
Fri Oct 17 07:32:33 PDT 2014







Dear all,

The last few days I also have been working on getting ASPECT to work on a mac. Following the comments on the bug (which I also encountered in the free surface cookbook when running it on multiple mpi processors) by Timo, I installed first the newest version of deal.ii and when that didn't solve the problem also the newest version of trilinos separately from the package installed by the dmg. All other dependencies like p4est, openmpi and petsc are still linked to this package. This didn't solve the problem either. Furthermore I have also tried it with older version of trilinos (stable release 11.10.1 and 11.8.1).  With PETSC it does work.

I must note that to be able to successfully install the newest git clone of trilinos I had to disable the packeges Claps, RBGen, Aristos and FEApp in PackagesList.cmake, because they could not be found. This was not the case in the mentioned stabled releases I tried.

To test if this is a problem with mac, or just the most recent version of trilinos, deal.ii and aspect, I installed it on a linux machine. It turned out this gave me the same problems. 

So my question is if the problem with trilinos and deal.ii is actually resolved or might I be doing something wrong with the installation (see settings below)?

Hope anyone can help with this!

Cheers,

Menno

parameters used for making trilinos:
cmake -D TrilinosFramework_ENABLE_MPI:BOOL=ON -D CMAKE_INSTALL_PREFIX:PATH=/Users/Menno/Documents/Phd/aspect/trilinos/build/ -D TPL_ENABLE_MPI:BOOL=ON -D BUILD_SHARED_LIBS:BOOL=ON -D CMAKE_BUILD_TYPE:STRING=RELEASE -D Trilinos_ENABLE_Fortran:BOOL=OFF -D Trilinos_WARNINGS_AS_ERRORS_FLAGS:STRING="" -D CMAKE_VERBOSE_MAKEFILE:BOOL=TRUE -D Trilinos_ENABLE_TESTS:BOOL=OFF -D Trilinos_ENABLE_ALL_PACKAGES:BOOL=OFF -D Trilinos_ENABLE_ALL_OPTIONAL_PACKAGES:BOOL=ON -D Trilinos_ENABLE_Epetra:BOOL=ON -D Trilinos_ENABLE_EpetraExt:BOOL=ON -D Trilinos_ENABLE_Tpetra:BOOL=ON -D Trilinos_ENABLE_Jpetra:BOOL=ON -D Trilinos_ENABLE_Kokkos:BOOL=ON -D Trilinos_ENABLE_Sacado:BOOL=ON -D Trilinos_ENABLE_Amesos:BOOL=ON -D Trilinos_ENABLE_AztecOO:BOOL=ON -D Trilinos_ENABLE_Ifpack:BOOL=ON -D Trilinos_ENABLE_Teuchos:BOOL=ON -D Trilinos_ENABLE_Rythmos:BOOL=ON -D Trilinos_ENABLE_Piro:BOOL=ON -D Trilinos_ENABLE_MOOCHO:BOOL=ON -D Trilinos_ENABLE_ML:BOOL=ON -D Trilinos_ENABLE_Thyra:BOOL=ON -D Trilinos_ENABLE_TrilinosCouplings:BOOL=ON $EXTRA_ARGS ../ 

The resulting output for two mpi processes (the more mpi processes are used, the faster the problem is encountered):
-----------------------------------------------------------------------------
-- This is ASPECT, the Advanced Solver for Problems in Earth's ConvecTion.
--     . version 1.2.pre
--     . running in DEBUG mode
--     . running with 2 MPI processes
--     . using Trilinos
-----------------------------------------------------------------------------

Number of active cells: 10240 (on 6 levels)
Number of degrees of freedom: 134692 (82818+10465+41409)

Number of free surface degrees of freedom: 20930
*** Timestep 0:  t=0 years
   Solving mesh velocity system... 0 iterations.
   Solving temperature system... 0 iterations.
   Rebuilding Stokes preconditioner...
   Solving Stokes system... 30+3 iterations.

   Postprocessing:
     Writing graphical output: output/solution-00000
     Topography min/max:       0 m, 0 m, 
     RMS, max velocity:        0.0109 m/year, 0.0297 m/year

*** Timestep 1:  t=5255.3 years
   Solving mesh velocity system... 9 iterations.
   Solving temperature system... 11 iterations.
   Rebuilding Stokes preconditioner...
   Solving Stokes system... 30+5 iterations.

   Postprocessing:
     Topography min/max: -53.33 m, 93.26 m, 
     RMS, max velocity:  0.00578 m/year, 0.0176 m/year

*** Timestep 2:  t=14136.7 years
   Solving mesh velocity system... 12 iterations.
   Solving temperature system... 11 iterations.
   Rebuilding Stokes preconditioner...
   Solving Stokes system... 30+5 iterations.

   Postprocessing:
     Topography min/max: -15.43 m, 36.03 m, 
     RMS, max velocity:  0.00694 m/year, 0.023 m/year

*** Timestep 3:  t=20932.2 years
   Solving mesh velocity system... 25 iterations.
   Solving temperature system... 11 iterations.
   Rebuilding Stokes preconditioner...
   Solving Stokes system... 30+4 iterations.

   Postprocessing:
     Topography min/max: -20.8 m, 59.68 m, 
     RMS, max velocity:  0.00612 m/year, 0.0212 m/year

*** Timestep 4:  t=28303.4 years
   Solving mesh velocity system... 90 iterations.
   Solving temperature system... 10 iterations.
   Rebuilding Stokes preconditioner...
   Solving Stokes system... 30+3 iterations.

   Postprocessing:
     Topography min/max: -21.81 m, 61.13 m, 
     RMS, max velocity:  0.00608 m/year, 0.0211 m/year

*** Timestep 5:  t=35711.1 years
   Solving mesh velocity system... 27 iterations.
   Solving temperature system... 10 iterations.
   Rebuilding Stokes preconditioner...
   Solving Stokes system... 30+1 iterations.

   Postprocessing:
     Topography min/max: -21.59 m, 61.11 m, 
     RMS, max velocity:  0.00609 m/year, 0.0211 m/year

*** Timestep 6:  t=43113.6 years


----------------------------------------------------
Exception on processing: 

--------------------------------------------------------
An error occurred in line <486> of file </Users/Menno/Documents/Phd/aspect/dealii/build/include/deal.II/lac/solver_cg.h> in

----- function
    void ---------dealii::SolverCG<de----------------------alii::Tr------------ilinosWrapper----
Exceptis::MPI::Vector>:on on processing: 
:solve(const MATRIX &, VECTOR &, const VECTOR &, const PRECONDI
----TIONER &) [VE------CTOR = de------------alii::TrilinosWrap------pers::MP----------I::Vector, MATRIX = -------------dealii::Tri-----
An error olinosWrappers::Sccurred iparseMatrn line <486> of filix, PRECONDITIONER = dealii::TrilinosWrappers::PreconditionAMG]
The violae </Users/Menno/Docted condition was: 
    false
The name uments/Phd/aspeand call sect/dealii/build/includequence of the exception was:/deal.II/lac/s
    SolverContrololver_cg::NoConvergen.h> in functionce (this->control()
    voi.last_step(),d dealii::Solve this->control().lasrCG<dealit_value())
Addi::TrilinosWraitional Information: 
ppers::MPIterative method reported convergeI::Vector>::solve(const Mnce failure in step 10ATRIX &, VECTOR &4650. The residual in the las, const Vt step was 4.25519e-19.

This errECTOR &, const PRECONDor messageITIONER &) [V can indicate that you have simpECTOR = dealii::Trilinoly not allowed a suffisWrappersciently large number ::MPI::Vector, MATRIX of iteratio= dealii:ns for your:TrilinosWrappers::SparseMa iterative solver to trix, PRECONDconverge. This often ITIONER = deahappens when you increalii::Trilise the size of ynosWraouppers:r prob:Precondilem. In tionAMG]such cas
The ves, the last reiolated conditionsidual  was: 
 will likely s   false
till be veThe name ry small, aand cnd you call sean make thequence of the exception was:
   error go away by i  SolverConncreasing trol::NoConvergence (this->control().last_step(), this->controlthe allowed number of iterations when setting up the SolverCont().last_varol object that determlue())
Additional Infoines the maximal number of irmation: 
Iterative methoterations yd reported conou allow.

Thvergencee other si failure in stetuation where this errp 104650. The residuaor may occur l in the last is when your mstep was 4.25519e-19.

atrix is not invertibleThis error (e.g., y message can inour matrix has dicate tha null-sat you have sipace), or if ymply not allowou try to appled a sufficiently large ny the wrong solver to a matrix (eumber of iterations for your ite.g., using CG for a matrrative solver to conix that is not symmetverric or not positive definite)ge. This often happ. In these caens whenses, the resi you incrdual in the laease the size of st iteration is likelyyour problem. In suc going to be large.
h cases, the last-------------------------- residual will l----------------------ikely --------still be v

Abortiery smallng!
------------, an----------------d you can make the--------------------- error go aw---
ay by increasing the allowed number of iterations when setting up the SolverControl object that determines the maximal number of iterations you allow.

The other situation where this error may occur is when your matrix is not invertible (e.g., your matrix has a null-space), or if you try to apply the wrong solver to a matrix (e.g., using CG for a matrix that is not symmetric or not positive definite). In these cases, the residual in the last iteration is likely going to be large.
--------------------------------------------------------

Aborting!
----------------------------------------------------
--------------------------------------------------------------------------
mpirun noticed that the job aborted, but has no info as to the process
that caused that situation.
--------------------------------------------------------------------------




> Date: Sat, 11 Oct 2014 19:18:45 +0200
> From: sylvia.rockel at fu-berlin.de
> To: aspect-devel at geodynamics.org
> Subject: Re: [aspect-devel] ASPECT on Mac
> 
> Dear all,
> 
> with PETSc instead of Trilinos the mpirun seems to work (step 480 now and
> still running). Thanks for the hint.
> 
> Concerning the Trilinos bug, I was totally aware of the fact that I'll
> have to wait until it's fixed in the mac package, since unfortunately it
> is Trilinos which I just can't manage to get it compiled without errors
> after I reset my whole computer. Wolfgang's reply was just to an older
> mail. So everything is fine. Until the Trilinos bug is fixed I can now use
> the Aspect_with_PETCs-version in the meantime.
> 
> But thanks anyways for your time and help.
> 
> Best,
> Sylvia
> 
> > Hey all,
> >
> > maybe I wasn't clear enough about this:
> >
> > The mac package deal.ii-8.2pre-v2 is broken because it bundles a
> > Trilinos version that is not compatible. As a result the code will
> > produce strange errors after a couple of timesteps in parallel
> > (basically the matrix is assembled wrong). You _have_ to wait until
> > Luca releases a new version or setup and compile deal.II by yourself.
> >
> >
> >> Since you have the deal.II package installed, you should also have
> >> PETSc.
> >> Can you try to compile ASPECT with PETSc support instead of using
> >> Trilinos?
> >> Maybe that gets you any further.
> >
> > Yes, that might work.
> >
> >
> > --
> > Timo Heister
> > http://www.math.clemson.edu/~heister/
> > _______________________________________________
> > Aspect-devel mailing list
> > Aspect-devel at geodynamics.org
> > http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
> >
> 
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel


 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/aspect-devel/attachments/20141017/9be99ff1/attachment.html>


More information about the Aspect-devel mailing list