PS: I should have clarified - I removed PETSC from the configure line of deal II because I wanted to nail the reason of my failure to run the deal examples to a single source of failure i.e., Trilinos. If I had Trilinos and PETSC enabled, I wouldn't have been able to single out the problem.
On 30 May 2011, at 11:18, Biswa Sengupta wrote: > Hi Toby, > > Yes, Trilinos 10.6.4 (mpi version) works fine, independent of deal II. I have > solved a couple of KKT systems using the linear and non-linear solvers. I had > also exported the library path using: > > export > DYLD_LIBRARY_PATH=$DYLD_LIBRARY_PATH:/Users/biswasengupta/Desktop/deal.II/lib > > Since I did not need PETSC, I have not installed it. ./configure doesn't > throw any error as well. So, I remain perplexed of what could be wrong. > > Best, > Biswa > > >> Biswa, >> >>> In my attempt to nail down the problem, I rebuilt Trilinos 10.6.4 >>> without MPI and then installed deal II without mpi support. In this >>> configuration the examples work as expected. So, in short I was >>> wondering if the mac users in the forum can let me know if they got >>> Trilinos 10.6.4 (MPI version) to work with deal II (MPI enabled). I >>> have tried both openMPI and MPICH2 compilers, but in vain. Both >>> compilers were built with gcc version 4.2.1 (Apple Inc. build 5666) >>> (dot 3). >> >> I don't know much about Trillinos; did you check he runs correctly on >> your machine *independently of deal.II*. If may simply be that the >> location of these libraries are still unknown to your system (see >> deal.II::FAQ for a clearer explanation of this). >> >> With PETSc you can do this already by (after compilation of PETSc) >> running "make test". That would confirm your PETSc (Trillinos) >> installation actually does work (ie linking with/without MPI). >> >> Do PETSc and Trilinos work by themselves? Can you compile, link, and >> run their test applications? >> >> Best, >> Toby >> >> -- >> Toby D. Young >> >> Assistant Professor Philosophy & Physics >> Polish Academy of Sciences >> www: >> http://www.ippt.gov.pl/~tyoung > > > On 30 May 2011, at 10:22, Biswa Sengupta wrote: > >> Dear all, >> >> In my attempt to nail down the problem, I rebuilt Trilinos 10.6.4 without >> MPI and then installed deal II without mpi support. In this configuration >> the examples work as expected. So, in short I was wondering if the mac users >> in the forum can let me know if they got Trilinos 10.6.4 (MPI version) to >> work with deal II (MPI enabled). I have tried both openMPI and MPICH2 >> compilers, but in vain. Both compilers were built with gcc version 4.2.1 >> (Apple Inc. build 5666) (dot 3). >> >> Any advice would be highly appreciated. >> >> Best, >> Biswa >> >> On 29 May 2011, at 17:15, Biswa Sengupta wrote: >> >>> Dear all, >>> >>> I am a brand new user of Deal II and after a lots of issues with compiler >>> flags, I have been able to compile the latest SVN version of Deal II with: >>> >>> ./configure --with-lapack --with-blas >>> --with-doxygen=/Applications/Doxygen.app/Contents/Resources/doxygen >>> -with-trilinos=/Users/biswasengupta/Desktop/tril CC=mpicc CXX=mpic++ >>> F77=mpif77 --disable-shared --with-umfpack --with-netcdf=/usr/local/netcdf >>> --with-metis=/Users/biswasengupta/Desktop/software_required/SuiteSparse/metis-4.0.3 >>> --enable-threads >>> >>> I have tried the compilation with and without PETSC, the error I describe >>> holds for both: >>> >>> 4565:step-5 biswasengupta$ make >>> Makefile:141: Makefile.dep: No such file or directory >>> ============================ Remaking Makefile.dep >>> ==============debug========= step-5.cc -> step-5.g.o >>> ============================ Linking step-5 >>> Undefined symbols: >>> "_Zoltan_Create", referenced from: >>> _ML_DecomposeGraph_with_Zoltan in libml.a(ml_agg_Zoltan.c.o) >>> "_Zoltan_Destroy", referenced from: >>> _ML_DecomposeGraph_with_Zoltan in libml.a(ml_agg_Zoltan.c.o) >>> "_Zoltan_Set_Param", referenced from: >>> _setup_zoltan in libml.a(ml_agg_Zoltan.c.o) >>> _setup_zoltan in libml.a(ml_agg_Zoltan.c.o) >>> _setup_zoltan in libml.a(ml_agg_Zoltan.c.o) >>> _setup_zoltan in libml.a(ml_agg_Zoltan.c.o) >>> _run_zoltan in libml.a(ml_agg_Zoltan.c.o) >>> _ML_DecomposeGraph_with_Zoltan in libml.a(ml_agg_Zoltan.c.o) >>> _ML_DecomposeGraph_with_Zoltan in libml.a(ml_agg_Zoltan.c.o) >>> _ML_DecomposeGraph_with_Zoltan in libml.a(ml_agg_Zoltan.c.o) >>> "_Zoltan_Set_Fn", referenced from: >>> _setup_zoltan in libml.a(ml_agg_Zoltan.c.o) >>> _setup_zoltan in libml.a(ml_agg_Zoltan.c.o) >>> _setup_zoltan in libml.a(ml_agg_Zoltan.c.o) >>> _setup_zoltan in libml.a(ml_agg_Zoltan.c.o) >>> "_Zoltan_Initialize", referenced from: >>> _ML_Operator_BlockPartition in libml.a(ml_op_utils.c.o) >>> _ML_DecomposeGraph_with_Zoltan in libml.a(ml_agg_Zoltan.c.o) >>> "_Zoltan_LB_Free_Part", referenced from: >>> _run_zoltan in libml.a(ml_agg_Zoltan.c.o) >>> _run_zoltan in libml.a(ml_agg_Zoltan.c.o) >>> "_Zoltan_LB_Partition", referenced from: >>> _run_zoltan in libml.a(ml_agg_Zoltan.c.o) >>> ld: symbol(s) not found >>> collect2: ld returned 1 exit status >>> make: *** [step-5] Error 1 >>> >>> >>> Now, if PETSC is enabled and I try to compile any of the examples, the >>> linker throws an error on the related petsc library as well. So, I guess >>> something fundamental is wrong. Is there any flag or log file where one can >>> check whether the compilation has truly been fault-less? Any suggestions >>> are welcome. BTW I am on a Mac with Snow Leopard (Dual Core). >>> >>> Best, >>> Biswa >> > > _______________________________________________ > dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
