Segmentation fault (core dumped) in HF calculation

[milesj]

Hello,

I'm trying to converge a Hartree Fock computation, but I keep running into a Segmentation fault error. I've attached the quantum espresso output files from the input that I used, as well as the failed yambo run. I've tried with a few different yambo compilations, playing with the linear algebra versions as indicated in this post viewtopic.php?t=1834, but the error remained the same.
I hadn't been getting this error before, with any BSE or GW calculations, but in this run I increased ecutwfc in my quantum espresso input. Presumably it's some sort of memory issue, but I've watched the nodes running the calculation and they never go anywhere near their memory limit (they don't even reach 1/5 the total ram on the node). My guess is that there's some sort of bug in how OpenMP is used, but I'm interested in your opinion. I'm using yambo-5.1.1

Thanks,
Miles

[Daniele Varsano]

Dear Miles,
can you please attach also the log files of your run?

Best,
Daniele

[milesj]

Attached, for one of the compilations it also has the line:

Code: Select all

*** stack smashing detected ***: <unknown> terminated

but for the most part the log files just say segmentation fault

[Daniele Varsano]

Dear Miles,
is this one of the log files contained in the LOG directory? It seems pretty much the batch script output.
I would like to have a look to the yambo log files to understand in which part of the run it crashes.

Best,
Daniele

[milesj]

My mistake, attached below.

[Daniele Varsano]

Dear Miles,

not totally clear what's happening, but it could be a memory issue.
Can you try to run using more CPUs (if you have them available) defining in input a parallelization scheme allowing memory distribution among cores:

Code: Select all

SE_CPU= "  1 1  #ncpu"               # [PARALLEL] CPUs for each role
 SE_ROLEs= "q qp b"  

where #ncpu us the total number of cores you are using.



Best,
Daniele