RI (Resolution of Identity) dies with no error, PBE0 calc

[DMT]

Hello,

I added

&RI T

&END RI

(all defaults)

section in the &HF section and cp2k dies with no error message at output:

 Extrapolation method: initial_guess


 SCF WAVEFUNCTION OPTIMIZATION

  Step     Update method      Time    Convergence         Total energy    Change
  ------------------------------------------------------------------------------

. On the terminal only the work "Killed" appears and all dies. No error messages in terminal, no error messages in output.

[Augustin Bussy]

This is likely a problem of available memory. Memory management in RI-HFX is not as robust. While the standard HFX scheme can run with any amount of available memory (at the cost of recomputing some integrals at each SCF cycle), RI-HFX requires enough to store all 3-center integrals at once.

From your output, I see that you have ~30GB of memory for a system with 936 basis functions. I also see that you are using 16 CPUs. I think that this is not enough to run such a large system at the HFX/hybrid-DFT level. At this point, my best advice would be to use a larger computer. Note that CP2K performance scales better with MPI ranks rather than OMP threads.

Best,

Augustin

---

From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of DMT <dobro...@gmail.com>
Sent: Thursday, December 22, 2022 1:50 PM
To: cp2k <cp...@googlegroups.com>
Subject: [CP2K:18246] RI (Resolution of Identity) dies with no error, PBE0 calc

 

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/42e3edcf-d290-4d0d-a4c5-34754007a2ffn%40googlegroups.com.

[DMT]

I was worried it was something like that ...

Regards,

Dobromir