; >>>>>>
>>>>>>> >>
>>>>>>>
>>>>>>> wrote:
>>>>>>>
>>>>>>> In my multilevel FETI-DP code, I have localized course matrices,
>>>>>>> which
>>&g
Thomas,
I'm missing one point...
You run N sequential factorizations (i.e. each has its own matrix to
work with and no need to communicate?) independently within ONE node? Or
there are N factorizations that run on N nodes?
Jed,
> MUMPS uses MPI_Iprobe on MPI_COMM_WORLD (hard-coded).
Any reas
I use a modified MPICH version. On the system I use for these
benchmarks I cannot use another MPI library.
I'm not fixed to MUMPS. Superlu_dist, for example, works also
perfectly for this. But there is still the following problem I cannot
solve: When I increase the number of coarse space mat
seems to be more involved.
Thanks,
-Nachiket
-- next part --
An HTML attachment was scrubbed...
URL:
<http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20121221/9a2a607d/attachment.html>
On Dec 21, 2012, at 3:16 PM, Nachiket Gokhale wrote:
> I have a dense matrix A (100x100) and I want to extract a matrix B from it
> consisting of the first N columns of A. Is there a better way to do it
> than getting the column using MatGetColumnVector, followed by VecGetArray,
> and
Dear all,
I am in the process of upgrading from 3.2 to 3.3.
I am a little bit puzzled by the following change:
> VecSetBlockSize() cannot be called after VecCreateSeq() or
> VecCreateMPI() and must be called before VecSetUp() or
> VecSetFromOptions() or before either VecSetType() or VecSetSizes()
On Fri, Dec 21, 2012 at 10:51 AM, Thomas Witkowski
wrote:
> I use a modified MPICH version. On the system I use for these benchmarks I
> cannot use another MPI library.
>
> I'm not fixed to MUMPS. Superlu_dist, for example, works also perfectly for
> this. But there is still the following problem
Okay, I did a similar benchmark now with PETSc's event logging:
UMFPACK
16p: Local solve 350 1.0 2.3025e+01 1.1 5.00e+04 1.0 0.0e+00
0.0e+00 7.0e+02 63 0 0 0 52 63 0 0 0 51 0
64p: Local solve 350 1.0 2.3208e+01 1.1 5.00e+04 1.0 0.0e+00
0.0e+00 7.0e+02 60 0 0 0
Dear all,
would it be possible to have a unified interface (also Fortran callable)
to the PetscKernel_A_gets_inverse_A_ routines?
I find them very useful within my own piece
of Fortran code to solve small dense linear system (which I have
to do very frequently).
I have my own interface, at present,
seeing.
-- next part --
An HTML attachment was scrubbed...
URL:
<http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20121221/92dbba70/attachment.html>
>>>> communicator, which is a subset of MPI::COMM_WORLD. The LU
>>>>>>> factorization of
>>>>>>> the matrices is computed with either MUMPS or superlu_dist, but both
>>>>>>> show
>>>>>>> some scaling property I really wonder of: When the overall problem
>>>>>>> size is
>>>>>>> increased, the solve with the LU factorization of the local matrices
>>>>>>> does
>>>>>>> not scale! But why not? I just increase the number of local
>>>>>>> matrices,
>>>>>>> but
>>>>>>> all of them are independent of each other. Some example: I use 64
>>>>>>> cores,
>>>>>>> each coarse matrix is spanned by 4 cores so there are 16 MPI
>>>>>>> communicators
>>>>>>> with 16 coarse space matrices. The problem need to solve 192 times
>>>>>>> with the
>>>>>>> coarse space systems, and this takes together 0.09 seconds. Now I
>>>>>>> increase
>>>>>>> the number of cores to 256, but let the local coarse space be
>>>>>>> defined
>>>>>>> again
>>>>>>> on only 4 cores. Again, 192 solutions with these coarse spaces are
>>>>>>> required, but now this takes 0.24 seconds. The same for 1024 cores,
>>>>>>> and we
>>>>>>> are at 1.7 seconds for the local coarse space solver!
>>>>>>>
>>>>>>> For me, this is a total mystery! Any idea how to explain, debug and
>>>>>>> eventually how to resolve this problem?
>>>>>>>
>>>>>>> Thomas
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>> --
>>>> 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/20121221/312cd24d/attachment-0001.html>
On Fri, Dec 21, 2012 at 7:04 AM, Aldo Bonfiglioli
wrote:
> Dear all,
> I am in the process of upgrading from 3.2 to 3.3.
>
> I am a little bit puzzled by the following change:
>> VecSetBlockSize() cannot be called after VecCreateSeq() or
>> VecCreateMPI() and must be called before VecSetUp() or
>
gt;> cores,
>>>>> each coarse matrix is spanned by 4 cores so there are 16 MPI
>>>>> communicators
>>>>> with 16 coarse space matrices. The problem need to solve 192 times
>>>>> with the
>>>>> coarse space systems, and this takes together 0.09 seconds. Now I
>>>>> increase
>>>>> the number of cores to 256, but let the local coarse space be defined
>>>>> again
>>>>> on only 4 cores. Again, 192 solutions with these coarse spaces are
>>>>> required, but now this takes 0.24 seconds. The same for 1024 cores,
>>>>> and we
>>>>> are at 1.7 seconds for the local coarse space solver!
>>>>>
>>>>> For me, this is a total mystery! Any idea how to explain, debug and
>>>>> eventually how to resolve this problem?
>>>>>
>>>>> Thomas
>>>>>
>>>>
>>>>
>>>>
>>>
>>
>> --
>> 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/20121221/8968e584/attachment-0001.html>
13 matches
Mail list logo
