HSE calculations stops with BUG report

[kdoblhoff]

I am running an HSE calculations. Some of the calculations stopped with the following report:

Code: Select all

 FFT: planning ... GRIDC
 FFT: planning ... GRID_SOFT
 -----------------------------------------------------------------------------
|                     _     ____    _    _    _____     _                     |
|                    | |   |  _ \  | |  | |  / ____|   | |                    |
|                    | |   | |_) | | |  | | | |  __    | |                    |
|                    |_|   |  _ <  | |  | | | | |_ |   |_|                    |
|                     _    | |_) | | |__| | | |__| |    _                     |
|                    (_)   |____/   \____/   \_____|   (_)                    |
|                                                                             |
|     internal error in: mkpoints_full.F  at line: 1170                       |
|                                                                             |
|     internal error in SET_INDPW_FULL: insufficient memory (see wave.F       |
|     safeguard) 541 540                                                      |
|                                                                             |
|     If you are not a developer, you should not encounter this problem.      |
|     Please submit a bug report.                                             |
|                                                                             |
 -----------------------------------------------------------------------------

 FFT: planning ... GRID
 -----------------------------------------------------------------------------
|                     _     ____    _    _    _____     _                     |
|                    | |   |  _ \  | |  | |  / ____|   | |                    |
|                    | |   | |_) | | |  | | | |  __    | |                    |
|                    |_|   |  _ <  | |  | | | | |_ |   |_|                    |
|                     _    | |_) | | |__| | | |__| |    _                     |
|                    (_)   |____/   \____/   \_____|   (_)                    |
|                                                                             |
|     internal error in: mkpoints_full.F  at line: 1170                       |
|                                                                             |
|     internal error in SET_INDPW_FULL: insufficient memory (see wave.F       |
|     safeguard) 541 540                                                      |
|                                                                             |
|     If you are not a developer, you should not encounter this problem.      |
|     Please submit a bug report.                                             |
|                                                                             |
 -----------------------------------------------------------------------------

I have not yet investigated in more detail what may have caused this. The only thing that I DO know is that other calculations (including calculations with the exact same setup but a k-point mesh of 7x7x1 and 8x8x1 instead of 6x6x1 DID run). I know that there have been issues with VASP on the cluster that I am using, but I believe that they are related to the use of vasp on the high memory nodes. I attach the input.

Code: vasp.6.3.0 20Jan22 (build Dec 15 2022 23:39:30) complex
toolchain used for compilation: foss-2021a
Executed on: 1 node, 32 out of 128 cpu and 256 GiB of memory in total (so 1/4 for use)

I would be grateful if someone could check whether this has to do with the particular compile or with the input.
Thank you and best regards,
Katharina

[alex]

Hello Katharina,

my first impression is that your memory restrictions in the INCAR for HSE and the 32 cores will not suffice.

Please compare the actual number of k-points plotted in the OUTCAR with the larger mesh (symmetry might help here).

Good luck!

alex

[fabien_tran1]

Hi,

By running the calculation using 32 cores I can indeed see that it runs with 5x5x1, 7x7x1 and 8x8x1, but not with 6x6x6. I suppose (?) that the distribution of the k-points can be done less efficiently with 6x6x1. However, using KPAR=4 in INCAR seems to be helpful for 6x6x1.