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

[pluto]

Dear Yichen,

A PdTe2 slab of 60 non-equivalent atoms needs up to 32 GB RAM in 4x 
k-point-parallel OMP=2 FM+SOC scf calculation with default settings and 
ecut 0.999 on modern i7 intel CPU. This is a very big slab, so for any 
typical case much less RAM is required.

On a cluster k-parallel works great if one can solve all the slurm 
issues.

I experience RAM leak in lapw1 when calculating bands for a huge mesh 
(in my case 2600 k-points) of k-points, but this can be fixed by 
granularity (as suggested by Prof. Blaha, depends on type of compiler). 
There is no RAM leak in lapwso. Again, 2600 k-points is not typical, 
typically one does maybe 100 k-points for bands.

Best,
Lukasz




-- 
PD Dr. Lukasz Plucinski
Group Leader, FZJ PGI-6
https://electronic-structure.fz-juelich.de/
Phone: +49 2461 61 6684


On 2024-06-26 00:20, Yichen Zhang wrote:
> Dear all,
> 
> Indeed this is a partial solution, which does not solve the essential
> stack limit problem! I do not know how to rebuild the kernel. I guess
> I’d rather not try or look for help on that at this point.
> It was fun getting WIEN2k to work on my PC, produce some simple bands,
> and learning about things :)
> For serious calculations, I will compile the programme on clusters,
> which shouldn't be a problem.
> 
> I’m not sure how much bigger systems my device would be able to
> handle. Naively, on the other hand, I expect that -fmax-stack-var-size
> limit just dumps all the larger variables to dynamic memory, which in
> principle is bounded by the machine itself. The device has a 64 GB
> unified memory, which should be able to hold capacity more than normal
> 64 GB RAM by roughly a factor of 2 (according to Apple :) if it’s
> anywhere reliable). But I will see how it does when it needs to deal
> with bigger systems at some point.
> 
> Best regards
> Yichen
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:
> http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html