[QE-users] Inability of QE compiled for GPUs to read SG15 ONCV Potentials

[Soham Ghosh]

Hello, 

While using ONCV potentials for a pw.x scf calculation, I noticed that QE compiled for GPUs (in this case 7.2 but I believe true for 7.3 as well) fails to read SG15 ONCV pseudopotentials. 
The same version of pw.x compiled for CPUs has no such problem reading the PPs ((once the value of index=“N”, N>9  that has been discussed before is fixed). 
I get the following output:

 end of file reached, tag PP_HEADER not found
 end of file reached, tag not found
 end of file reached, tag PP_MESH not found
 end of file reached, closing tag not found
…
           0            1

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     Error in routine set_dft_from_name (2):
      conflicting values for iexch
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

The Pseudopotentials are obtained from http://www.quantum-simulation.org/potentials/sg15_oncv/

I see it for both scalar relativistic and fully relativistic pseudo potentials. The specific pseudpotentials for which it failed are
Br_ONCV_PBE-1.2.upf
Br_ONCV_PBE_FR-1.0.upf 

Could you please let me know if you have any idea why? I will be happy to share the build info if you need it. 

Thanks, 
Soham Ghosh
HPC performance engineer,
NERSC at LBL
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20240425/75b78643/attachment.html>