[QE-users] [QE-GPU] pDOS abruptly ends before Fermi Energy

[Paolo Giannozzi]

On 11/07/2025 23:25, André Luiz Koch Liston wrote:

> Has this issue been seen before? What steps may I take to troubleshoot this?

a similar weird problem with projwfc.x on GPUs has been recently 
reported directly to me. It consists in NaN's partially filling the 
output, is observed only for large calculations and only in some cases 
with no apparent logic. No such problem on CPUs. No clues.

There isn't much that can be done in similar cases. The code seems to 
stop before completing. As a first step, you should figure out where it 
stops (assuming it always stops in the same place, which may or may not 
happen). A not-so-smart technique is to put some "write" here and there, 
ensuring that what is written is immediately flushed to output

Paolo
-- 
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 206, 33100 Udine Italy, +39-0432-558216