[petsc-users] Compiling PETSc 3.4.3 with ICC 14.0 and OpenMPI 1.6.5

[Matthew Knepley]

On Wed, Oct 30, 2013 at 4:48 PM, Maxime Boissonneault <
maxime.boissonneault at calculquebec.ca> wrote:

>
> Le 2013-10-30 17:45, Matthew Knepley a écrit :
>
> On Wed, Oct 30, 2013 at 4:41 PM, Maxime Boissonneault <
> maxime.boissonneault at calculquebec.ca> wrote:
>
>>  Hi,
>> Here is the attached configure.log file.
>>
>> I will try the flag --with-debugging tomorrow.
>>
>
>  Get rid of -O2: CFLAGS="-O2 -xHost -mkl -fPIC -m64" FFLAGS="-O2 -xHost
> -mkl -fPIC -m64" --prefix=/software6/libs/petsc/3.4.3_intel
>
> I had -O3 earlier. The error was the same. It is overwritten by the
> makefile anyway if you look in the details. This is not the source of the
> problem.
>

I know, but if you leave it there with you use --with-debugging, it could
still cause the problem.


>   --with-dynamic-loading
> --with-x=0 --with-mpi-compilers=1
> --with-mpi-dir=/software6/mpi/openmpi/1.6.5_intel
>  --known-mpi-shared-libraries=1
> --with-debugging=no
> --with-shared-libraries=1
>
> --with-blas-lapack-dir=/software6/compilers/intel/composer_xe_2013_sp1/lib/intel64
> Don't need this --with-scalapack=1
>
> --with-scalapack-include=/software6/compilers/intel/composer_xe_2013_sp1/include
>
> --with-scalapack-lib=/software6/compilers/intel/composer_xe_2013_sp1/lib/intel64/libmkl_scalapack_lp64.a
> Don't need this --with-blacs=1
>
> --with-blacs-include=/software6/compilers/intel/composer_xe_2013_sp1/include
>
> --with-blacs-lib=/software6/compilers/intel/composer_xe_2013_sp1/lib/intel64/libmkl_blacs_openmpi_lp64.a
> --download-mumps=yes
> --download-ptscotch=1
> --download-superlu=yes
> --download-superlu_dist=yes
>  --download-parmetis=yes
> --download-metis=yes
> --download-ml=yes
> This is deprecated. Use -pc_type gamg --download-prometheus=yes
>
> Which part is, download-ml ? or download-hypre ? Again, this is not the
> module that fails compiling however.
>

I do not understand the question here. ML is algebraic multigrid, similar
to -pc_type gamg. Hypre has a few things in it.

  Matt


> Maxime
>
>  --download-hypre=yes
>
>    Matt
>
>>
>>  Maxime
>>
>> Le 2013-10-30 16:49, Matthew Knepley a écrit :
>>
>> On Wed, Oct 30, 2013 at 3:39 PM, Maxime Boissonneault <
>> maxime.boissonneault at calculquebec.ca> wrote:
>>
>>>  Hi,
>>> I am currently trying to compile PETSc 3.4.3 with ICC 14.0 and OpenMPI
>>> 1.6.5 and it fails SuperLU_Dist.
>>>
>>> I am getting the following error :
>>>
>>> 68303 /software6/mpi/openmpi/1.6.5_intel/bin/mpicc
>>> -I/software6/mpi/openmpi/1.6.5_intel/include
>>> -I/software6/mpi/openmpi/1.6.5_intel/include -O2 -xHost -mkl -fPIC -m64
>>> -fPIC -O3  -DAdd_ -DUSE_VENDOR_BLAS -c pdgstrf.c
>>> 68304 make[1]: Leaving directory
>>> `/software6/src/petsc-3.4.3-intel-try/externalpackages/SuperLU_DIST_3.3/SRC'
>>> 68305 error #13002: unexpected CFE message argument:  e. The staggered
>>> cosine transform may be
>>> 68306 warning #13003: message verification failed for: 556; reverting to
>>> internal message
>>> 68307 pdgstrf.c(2672): warning #556: a value of type "int" cannot be
>>> assigned to an entity of type "MPI_Request"
>>> 68308             U_diag_blk_send_req[krow] = 1; /* flag outstanding
>>> Isend */
>>> 68309                                       ^
>>> 68310
>>> 68311 pdgstrf.c(2672): warning #152: Fatal error: Trigonometric
>>> Transform has failed to release the memory.
>>> 68312             U_diag_blk_send_req[krow] = 1; /* flag outstanding
>>> Isend */
>>> 68313                                         ^
>>> 68314
>>> 68315 compilation aborted for pdgstrf.c (code 1)
>>> 68316 make[1]: *** [pdgstrf.o] Error 1
>>>
>>> Anyone has a clue on to what might be going on ?
>>>
>>
>>  1) Always send configure.log
>>
>>  2) It looks like an Intel compiler bug. Try building with
>> --with-debugging first
>>
>>    Matt
>>
>>
>>>
>>> --
>>> ---------------------------------
>>> Maxime Boissonneault
>>> Analyste de calcul - Calcul Québec, Université Laval
>>> Ph. D. en physique
>>>
>>>
>>
>>
>>  --
>> What most experimenters take for granted before they begin their
>> experiments is infinitely more interesting than any results to which their
>> experiments lead.
>> -- Norbert Wiener
>>
>>
>>
>> --
>> ---------------------------------
>> Maxime Boissonneault
>> Analyste de calcul - Calcul Québec, Université Laval
>> Ph. D. en physique
>>
>>
>
>
>  --
> What most experimenters take for granted before they begin their
> experiments is infinitely more interesting than any results to which their
> experiments lead.
> -- Norbert Wiener
>
>
>
> --
> ---------------------------------
> Maxime Boissonneault
> Analyste de calcul - Calcul Québec, Université Laval
> Ph. D. en physique
>
>


-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20131030/9d4a58dd/attachment.html>