BUG with ultra-large system

[Jianguo Si]

Dear developers,

When performing a VASP calculation on a relatively large system (approximately 200 atoms and 1,500 electrons), the convergence calculations were completed successfully, and the WAVECAR and CHGCAR files were also written successfully. However, the calculation automatically terminates with an error when writing the PROCAR file. I checked the log, which is shown below. Could you please help me understand what happened? It seems like it might be a bug? How to fix it?

-----------------------------------------------------------------------------
| _ ____ _ _ _____ _ |
| | | | _ \ | | | | / ____| | | |
| | | | |_) | | | | | | | __ | | |
| |_| | _ < | | | | | | |_ | |_| |
| _ | |_) | | |__| | | |__| | _ |
| (_) |____/ \____/ \_____| (_) |
| |
| internal error in: mpi.F at line: 1825 |
| |
| M_sumb_d: invalid vector size n -1350075136 |
| |
| If you are not a developer, you should not encounter this problem. |
| Please submit a bug report. |
| |
-----------------------------------------------------------------------------

Best Regards,
Jianguo Si

[Jianguo Si]

See the attchment.

[Jianguo Si]

Files related to the calculations are also sent as attachments

[andreas.singraber]

Dear Jianguo Si,

thank you for your report and for uploading the relevant files! The error message suggests that there is an integer variable overflowing while computing the size of to-be-communicated data (some product of integers exceeds the limit of 2147483647 for 4-byte integers). I looked into the source responsible for writing the PROCAR file and indeed there are two problematic M_sum_d procedure calls which can trigger an overflow in VASP 6.3.0. Fortunately, these issues were already found before and they should be fixed in VASP 6.4.0 and higher. Can you try upgrading to a newer version of VASP and check if the error persists?

All the best,
Andreas Singraber