[Wien] lapw0 stuck/drained with seemingly no error message upon launching SCF

[Yichen Zhang]

Dear Lukasz,

Thank you very much for sharing the experience and the comments.

I’m sorry that I made some mistakes in previous threads due to some of my misconceptions. I thought -fmax-stack-var-size=n moves local variables>n from stack to heap memory, but upon reading gfortran 11 documentation (https://devdocs.io/gnu_fortran~11/code-gen-options#Code-Gen-Options), I realized that they actually go to static memory when > n.

It seems slightly worse than a memory leak of heap memory since they are allocated from the beginning and similarly don’t get deallocated.

It is helpful for me to know that granularity settings can help with memory leak in lapw1 when running on heterogeneous environment and there is no leak in lapwso. Thank you.

So far, I’ve only been running wien2k on normal or typically small systems, without supercell and so on, on my PC Mac, so haven’t encountered memory issues again, with a 64 GB RAM.

It is good that there are many threads in the mailing list here addressing problems of running on a slurm cluster. It will be helpful for me when I need to run something on our slurm clusters.

Best regards
Yichen
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20240706/c7869df1/attachment.htm>