Positron Doppler calculation fails after SCF steps

Running Abinit Ground state
1

Hai all
Nice to work with latest v9.4.2.
But I have problem for positron calculations with v9.4.2.

system is W super cell with 127 atoms + 1 vacancy.
relaxation with positron and positron lifetime calculations are fine.
But doppler calculation fails after completing both electron and positron SCFs
please see attached log and input files

Mem: 132124428k total shared by 24cpus

ps: I am able to calculate doppler for 16 atoms super cell.
currently trying with 8cpu run to see if failure is due to memory requirement

looking for some clues to move forward
thanks
-rajaraman

W_444_1V_unrelaxed_db.abi (9.1 KB)

W_444_1V_unrelaxed_db.log (496.0 KB)

2

Hello Rajaraman,

Any updates on your memory tests?
What happens if you set your stack limit to unlimited while maintaining your ~130GB of memory requested for the job, i.e. including this line in your job submission file:

ulimit -s unlimited

?

Best,
Bogdan

3

Dear Bogdan,
Thanks for your suggestion.
I added ulimit line to the job script.

will get back with its outcome.

thanks
-rajaraman

4

Dear Bogdan,

Thanks a lot for the suggestion to use
ulimit -s unlimited

it worked.
though 39 SCFs took about 9hrs, Doppler calculation took additional 67hrs!
see attached log.

db.log (918.5 KB)

There was another bottleneck.
I needed optforces switch “on” for smooth switching from electronic SCFs to positron SCFs.
optforces 1

without “optforce” job fails at the start of positron SCFs
see failed job log.

db-fail_no_optforce.log (375.9 KB)

though it is ok as long as jobs complete, but the optforce is redundant when we do doppler calculations
Positron - abinit (tpositron_5 or 6)

thanks
-rajaraman

5

Dear Rajaraman,

I am glad this solution worked fine for you.

I suspect it might still be a memory related problem, but I will have to ask someone well-versed in this aspect.

For the future, I might suggest to add -fbacktrace to your FLAGS when compiling ABINIT because at this stage I can’t reproduce your error.
Please let me know of the outcome if you pursue this.

Best,
Bogdan

6

Thank you Bogdan.

I am about to recompile as the output of the completed run showed only NAN.

I will include “backtrace” flag now.

I dont see any such switch to add -fbacktrace.

how to put it during compilation?

thanks
-rajaraman