[Wannier] parallel computing for plotting MLWFs

[Giovanni Pizzi]

Dear Shima,

Indeed, the plotting routines have not been parallelized yet so they will run only in serial.
Parallelizing them should be quite straightforward, so anyone willing to prepare a pull request is welcome to do so on the GitHub page<https://github.com/wannier-developers/wannier90>.

Regarding the interpolation of the band structure, you could use the `geninterp` module of postw90.x is instead parallelized - you can find instructions on how to use it in the documentation (http://www.wannier.org/support/, User Guide, page 159.
For parallel efficiency, I suggest you investigate (check the timings) of the use of geninterp_single_file = .false. vs the default geninterp_single_file = .true. (described in the documentation) that will give important performance improvements for small systems (but for a big system it might be less relevant).

Best,
Giovanni Pizzi

P.S.: don't forget to sign your posts with your full name and affiliation

--
Giovanni Pizzi
EPFL and NCCR MARVEL
http://people.epfl.ch/giovanni.pizzi
http://nccr-marvel.ch/en/people/profile/giovanni-pizzi

On 2 May 2021, at 11:21, Shima Sharifi <shima.sharifi at tyndall.ie<mailto:shima.sharifi at tyndall.ie>> wrote:

Dear Wannier90 users,
I was reading the recent paper by authors of Wannier90 code: "Wannier90 as a community code: new features and applications". In section 3.4 they talk about the parallelisation scheme for the calculation of MLWFs using MPI (Minimisation of the spread and so on).
For the calculation that I am doing using this code (with the interface from quantum Espresso 6.4.1) , these steps are quite fast. Actually, the most time consuming part is plotting MLWFs over a fine real space grid. I need to calculate MLWFs within large supercells made of 10*10*10 primitive unit cells. Then, I'm interested in running parallel calculation for this part, or interpolating the band structure or plotting Fermi surfaces.
Then, my question is that if Wannier90 v. 3.0.0 has this ability to be run in parallel in order to calculate these properties,
Or at the moment, one is just able to do disentanglement and Wannierisation processes with MPI?

Thank you in advance.

Regards,
Shima
_______________________________________________
Wannier mailing list
Wannier at lists.quantum-espresso.org<mailto:Wannier at lists.quantum-espresso.org>
https://lists.quantum-espresso.org/mailman/listinfo/wannier

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20210503/d70f35f8/attachment.html>