[QE-developers] CUDA routine development for the TDDFT module

[O. Baris Malcioglu]

Dear all,

I recently got accepted to a permanent position in METU, and I am no longer
obliged to use other package programs.

In the meantime, I have developed a GPU/CPU FEM/FET based multi scale
charge state correction scheme which is able to handle atomic to meso
scales simultaneously, thanks to adaptive meshing  techniques and some
clever tricks I learned from the fenics project. The project was presented
in DFG and PSI-K meetings, and it is almost mature enough to publish.

An experimenter friend in Germany asked me a question regarding multi-scale
consequences of  optical anistropy in layered materials. I intend to use
(no longer active) frequency dependent polarizability calculation scheme I
implemented in TDDFT module years back in order to extend the capabilities
of the above program to see if I can make a feasible approximation. I
think, it will still be a handy tool for computer aided material research
for fiber optics even if it doesn't address adequately this particular
question.

In order to do this, I intend to develop some CUDA routines (10+, since I
really need the memory provided by the new link scheme) for the TDDFT
module.

Is it possible for me to join the current TDDFT team for this, or is it
better for me to fork the TDDFT and work on my own?

Please note that, it will take some time until my teaching workload eases,
my grant proposals are answered and I have some students. So it will be a
(very) slow progress for at least a year.

All the best,
Baris
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/developers/attachments/20181219/35d50b70/attachment.html>