Hello Jinhyun
This is not an error. Since you have discontinuous solution, at a face in 2d
you have two different values and at a vertex in 2d you may have 4 different
values. When deal.II saves solution files for visualization, it does not do any
averaging etc. at face/vertex, hence you will
Hello,
While solving a simple problem using FE_Q_DG0 elements, I found that the
output data vector (created by DataOut.add_data_vector) has non-unique
values for the same point. Attached is an example .pvd file. Here, there
are two points of same coordinates (0.1,0.95,0), and the pressure
Thank you for this tip. For my need, I only require the starting global index
on each partition since I am using local_element to do some operations like
multiplying by a local time step which is different for each cell. The
local_range function gave this info. Here is a sample usage in my code
On 2/1/19 7:19 AM, gabriel.pet...@koeln.de wrote:
> I tried some things, but also if I insert a safety factor of 0.1 or less
> the GMRES-procedure in the diffusion-step diverges.
> I implemented the time step function such that
>
> dt <= max_{T} h_T/(u_max(T))
>
> where the max is taken over
I tried some things, but also if I insert a safety factor of 0.1 or less
the GMRES-procedure in the diffusion-step diverges.
I implemented the time step function such that
dt <= max_{T} h_T/(u_max(T))
where the max is taken over all cells and u_max is the maximal value of u
on the cell T.
Praveen,
You can get the locally only owned IndexSet
https://dealii.org/developer/doxygen/deal.II/classLinearAlgebra_1_1distributed_1_1Vector.html#aff5265a5af79342c77ac50d584c9d3e4
and ask for the iterator
Maxi,
[...]
> Apparently nvcc does not know the intrinsics, and thus is throwing an
> error. Everything works fine when disabling the cuda part of deal.II
>
Yes, we disabled vectorization when compiling CUDA device code in
https://github.com/dealii/dealii/pull/7655 in a more fine-grained way.
It looks like as if this compilation line is responsible:
cd ~/Downloads_ICC/dealii/build/source/lac && /share/apps/software/Core/CUDA
/10.0.130/bin/nvcc -ccbin=/share/apps/software/Compiler/intel/2018.3.222-GCC
-7.3.0-2.30/impi/2018.3.222/bin64/mpiicpc -I~/Downloads_ICC/dealii/build/
source/lac
I removed those files, and rebuild using -xhost, but I still get the same
errors. AVX should be available, after I am compiling on a Xeon E5-2695 v4,
which definitely has AVX (and the intel compiler should have that, too).
Test programs using those functions work fine.
Am Freitag, 1. Februar
Dear Maxi,
I have not yet look at all details, but what strikes my attention is
that you added `-march=native` to your C++ compile flags. That is a gcc
thing, so I don't know what Intel makes out of that. Can you try what
happens if you replace it by `-xhost` which is (mostly) the Intel
I tried to compile deal.II using ICC 18, but during compilation I got
several errors, such as
~/Downloads_ICC/dealii/include/deal.II/base/vectorization.h(1489): error: no
operator "+=" matches these operands
operand types are: __m256d += const __m256d
11 matches
Mail list logo