A Segmentation fault when calculating the vacancy energy of 256 iron atoms
Describe the bug
When calculating the vacancy energy of 256 (actually 255) iron atoms by using 64 cpus and 2 processors with ABACUS(v3.7.3) , there is a Segmentation fault.
And I tried 28 cpus and 4 processors, it can run.
(1)64 cpus & 2 processors :
(2)28 cpus & 4 processors:
Expected behavior
What's the relationship between calculation success and the number of cpus and the number of processes.
To Reproduce
Environment
No response
Additional Context
No response
Task list for Issue attackers (only for developers)
- [ ] Verify the issue is not a duplicate.
- [ ] Describe the bug.
- [ ] Steps to reproduce.
- [ ] Expected behavior.
- [ ] Error message.
- [ ] Environment details.
- [ ] Additional context.
- [ ] Assign a priority level (low, medium, high, urgent).
- [ ] Assign the issue to a team member.
- [ ] Label the issue with relevant tags.
- [ ] Identify possible related issues.
- [ ] Create a unit test or automated test to reproduce the bug (if applicable).
- [ ] Fix the bug.
- [ ] Test the fix.
- [ ] Update documentation (if necessary).
- [ ] Close the issue and inform the reporter (if applicable).
I think there must be something went wrong in 3.7.x version, recently i also met problems but SIGNAL 9 AND 6.
I suggest you to try version 3.6.5 in https://github.com/deepmodeling/abacus-develop/releases/tag/v3.6.5
Reach me if it works.
I tired v.3.6.4 before. There was the same Segmentation.
@xdzhu
Is it OK for 1 mpi case?
For 1 processor, it still has a Segmentation fault.
There seems to be some bug with ELPA, could you please try ks_solver scalapack_gvx?
Setting ks_solver= scalapack_gvx doesn't help.
Can you try the latest version 3.7.5 to check whether it's solved or not?
with elpa
Now this bug is solved in abacus v3.8.0.