Wrong spectra for some BSE solvers

You can find here problems arising when using old releases of Yambo (< 5.0). Issues as parallelization strategy, performance issues and other technical aspects.

Moderators: Davide Sangalli, andrea.ferretti, myrta gruning, andrea marini, Daniele Varsano, Conor Hogan

Locked
Bruno
Posts: 72
Joined: Tue Dec 08, 2020 11:16 am

Wrong spectra for some BSE solvers

Post by Bruno » Thu Mar 17, 2022 10:29 am

Hello,

I'm facing a quite weird problem during the BSE solving. I can get a nice and meaningful BSE spectra if I use the haydock solver, both for the bse and ip spectra that are printed. But if I try to use diagonalization or inversion I get a weird bse spectra, but the IP part still looks fine. I'm not sure exactly what is the cause of that since the haydock works. Note that the weird spectra that I get between inversion and diagonalization is identical. Is it possible that I have a compilation problem?

I'm attaching the files 'inv.dat', 'diag.dat' and 'hay.dat' for the inversion, diagonalization and haydock o- outputs from yambo. All calculations were made on top of the same SAVE/ folder just copied for different location, and except for the solver all the parameters should be the same.
Archive.zip
Regards,
You do not have the required permissions to view the files attached to this post.
MSc. Bruno Cucco
PhD Candidate
CNRS Institut des Sciences Chimiques de Rennes, France
Université de Rennes 1
https://iscr.univ-rennes1.fr

User avatar
Davide Sangalli
Posts: 610
Joined: Tue May 29, 2012 4:49 pm
Location: Via Salaria Km 29.3, CP 10, 00016, Monterotondo Stazione, Italy
Contact:

Re: Wrong spectra for some BSE solvers

Post by Davide Sangalli » Thu Mar 17, 2022 11:49 am

Dear Bruno,
my guess is that there is something wrong with the lapack / scalapack linear algebra subroutines that are used both by the diago and by the inversion solver.

Can you try one of these two options (or even both)
(i) to recompile changing the linked libraries
(ii) to use the slepc solver

Best,
D.
Davide Sangalli, PhD
CNR-ISM, Division of Ultrafast Processes in Materials (FLASHit) and MaX Centre
https://sites.google.com/view/davidesangalli
http://www.max-centre.eu/

Bruno
Posts: 72
Joined: Tue Dec 08, 2020 11:16 am

Re: Wrong spectra for some BSE solvers

Post by Bruno » Mon Mar 21, 2022 3:42 pm

Hello Davide,

As always, thanks for your help. It really seems related to the libraries but I will have to do some tests and recompile yambo again to be sure, unfortunately the cluster was recently updated and lot has changed regarding libraries path and so on, and also it's a Cray based system. I'll ask their assistance to recompile it. For now I'm just using it on another cluster so it's not extremely urgent.

Regards,
MSc. Bruno Cucco
PhD Candidate
CNRS Institut des Sciences Chimiques de Rennes, France
Université de Rennes 1
https://iscr.univ-rennes1.fr

Locked