anton:
Sounds good, 11?
Sounds good, 11?
basnijholt:
@cwg @xavier, I made a start with the conda-recipe for
@cwg @xavier, I made a start with the conda-recipe for
- tkwant: conda-forge/staged-recipes#12116
- kwantspectrum: conda-forge/staged-recipes#12117
xavier:
There is a tkwant channel BTW.
There is a tkwant channel BTW.
michaelwimmer:
yes
yes
slavoutich:
I am still chatting in the coffee room. Should we zoom?
I am still chatting in the coffee room. Should we zoom?
anton:
Yep, link above
Yep, link above
cosma:
I noticed that kwant.operator.Current does not live inside a .py file, but in something that looks precompiled. was it written in python and then cythoned? is there a way to see the code itself? I'm asking because I want to compute current operators in non-Hermitian systems, so I might have to go around the pre-built functionality of kwant
I noticed that kwant.operator.Current does not live inside a .py file, but in something that looks precompiled. was it written in python and then cythoned? is there a way to see the code itself? I'm asking because I want to compute current operators in non-Hermitian systems, so I might have to go around the pre-built functionality of kwant
cwg:
The module
The module
kwant.operator is written in Cython. The sources are of course available (git or tarball).
cwg:
Cosma, did you check out the
Cosma, did you check out the
check_hermiticity flag of the existing operator classes?
cosma:
thx
thx
cwg:
Perhaps this will do just what you need?
Perhaps this will do just what you need?
cosma:
this is very interesting... I think my problem is that I have a model in which hoppings t_ab > t_ba, so I don't build it directly using kwant
instead, I get the Hamiltonian by summing up its Hermitian and anti-Hermitian parts (both of which are obtained using kwant)
I'm not sure if check_hermiticity will help me... if kwant.operator.Current(syst) uses the hopping functions of syst, then probably not
this is very interesting... I think my problem is that I have a model in which hoppings t_ab > t_ba, so I don't build it directly using kwant
instead, I get the Hamiltonian by summing up its Hermitian and anti-Hermitian parts (both of which are obtained using kwant)
I'm not sure if check_hermiticity will help me... if kwant.operator.Current(syst) uses the hopping functions of syst, then probably not
cwg:
So by doing that and using the built-in operators with
So by doing that and using the built-in operators with
check_hermiticity=False you might be able to do what you want using entirely built-in Kwant functionality.
anton:
Cosma has a nonhermitian hopping, that wouldn't work.
Cosma has a nonhermitian hopping, that wouldn't work.
slavoutich:
This commit attracted my attention:
https://gitlab.kwant-project.org/kwant/kwant/-/commit/4fd146cdec4a5f7daac974d308086e9e9633ad0c
Yeah, it inserts a bunch of ^L in builder.py. But why?
anton:
I think @cwg uses these.
I think @cwg uses these.
...when trying to install, Ubuntu 20.04 reports that "The repository 'http://ppa.launchpad.net/kwant-project/ppa/ubuntu focal Release' does not have a Release file".
anton:
No, you need to:
- Install miniconda
- Execute
conda install kwant -c conda-forge
many thanks, I'll try it.
Hi all, I am here again. I have also a question about parallelization. For instance, when this line "trans = smatrix.transmission(1, 0)" is executing, is kwant executing it in parallel?. Suppose I am running it in a 4-core processor. Does kwant use the whole available cores, or just only one?.
slavoutich:
Is there a code style guide for Kwant somewhere, except https://kwant-project.org/contribute? Like, git conventions are rather strict and not mentioned here :)
Is there a code style guide for Kwant somewhere, except https://kwant-project.org/contribute? Like, git conventions are rather strict and not mentioned here :)
anton:
Not as far as I know, but please add.
Not as far as I know, but please add.
slavoutich:
Also, it might be cool to add
Also, it might be cool to add
.pylintrc to the project.
anton:
Sure
Sure
jbweston:
It should use only 1 core. If you see more than 1 core working it's because one of the low-level linear algebra libraries is trying to use >1 core, but typically this actually makes Kwant calculations run slower.
You can set the environment variable OMP_NUM_THREADS=1 and that should ensure that everything runs on 1 core.
anton:
(sorry, this morning didn't work out)
(sorry, this morning didn't work out)
slavoutich:
👍
👍
anton:
Starting in a minute
Starting in a minute