Page 2 of 2
Re: internal error in SET_INDPW_FULL: insufficient memory
Posted: Tue Jun 04, 2024 1:27 am
by sergey_lisenkov1
Not yet. I'll try it, however switching off symmetry especially in hybrid functional calculations will certainly increase a computational time.
Re: internal error in SET_INDPW_FULL: insufficient memory
Posted: Tue Jun 04, 2024 7:29 am
by henrique_miranda
I agree. But since you don't have many symmetries, perhaps the execution time difference is not too big.
We will try to fix this in the code, but it will take longer.
By turning off symmetries, you might be able to run the calculation now.
Re: internal error in SET_INDPW_FULL: insufficient memory
Posted: Wed Jun 05, 2024 2:36 pm
by sergey_lisenkov1
I'm not sure that switching off symmetry is a good way to go - in my hybrid functional calculations it will dramatically increase number of k-points which will lead to very long computation time.
Re: internal error in SET_INDPW_FULL: insufficient memory
Posted: Wed Jun 05, 2024 3:24 pm
by henrique_miranda
There are many ways you can circumvent this issue:
1. You can try and lower NCORE. Currently, you have NCORE=20, I tried NCORE=10 for example and it works. You should probably set it lower to get better performance in hybrid calculations (you need to do some testing on your own hardware).
2. I symmetrized (a better word for it is standardized) your structure with a python script similar to the one I indicated you and I get this POSCAR:
Code: Select all
symmetrized poscar
1.0
3.7447066163850500 -3.7522035265279627 0.0000000000000000
3.7447066163850500 3.7522035265279627 0.0000000000000000
0.0000000000000000 0.0000000000000000 5.3011160000000004
Cs Ge Cl
1 1 3
Direct
0.5000000000000000 0.5000000000000000 0.5000000000000000
0.0000000000000000 0.0000000000000000 0.0000000000000000
0.5000000000000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 0.5000000000000000 0.0000000000000000
0.0000000000000000 0.0000000000000000 0.5000000000000000
It keeps the distortion you applied and it has the same point group.
Using it, the calculation does not stop anymore.
3. Slightly tuning ENCUT from 544.23 in your INCAR file to for example 544 also works.
4. Turning off symmetries (ISYM=0) also works. In your calculation you have 27 kpoints in the IBZ, with ISYM=0 you get 63 kpoints. A factor of 2 in your calculation time.