Could you: 0) Send the FORCE_STRESS file? 1) Use siesta-4.1 b4 (you are using b3) 2) Add MinSCFIterations 3 to the options 3) I would probably try and relax the system a bit first in a non-polarized calculation (but this is unrelated to the failure)
Also, you are compiling with Intel -O3 which I typically won't advice. Intel compilers are extremely aggressive which may lead to small numerical problems. I would stick with -O2 and compile atom.F with -O1. It seems the output was truncated since the failing call is *after* the atoms has been moved. I.e. try this: $> unbuffer mpirun -np ... siesta RUN.fdf > RUN.out or equivalent, then send the output again. Den man. 15. apr. 2019 kl. 22.02 skrev I. Camps <ica...@gmail.com>: > Hi Nick, > > Here it is. > > []'s, > > Camps > > > On Sun, Apr 14, 2019 at 5:02 PM Nick Papior <nickpap...@gmail.com> wrote: > >> Could you please do a tar.gz or zip file? rar is horrible ;) >> >> Den lør. 13. apr. 2019 kl. 22.01 skrev I. Camps <ica...@gmail.com>: >> >>> Hello Nick, >>> >>> Here are the files. >>> >>> []'s, >>> >>> Camps >>> >>> >>> On Fri, Apr 12, 2019 at 5:02 PM Nick Papior <nickpap...@gmail.com> >>> wrote: >>> >>>> Could you attach the output, fdf and psf files? >>>> >>>> Den tor. 11. apr. 2019 kl. 22.01 skrev I. Camps <ica...@gmail.com>: >>>> >>>>> Hello, >>>>> >>>>> My system is a nanotube with 123 atoms (B, N, Ni). I am running a >>>>> geometry optimization and the calculations are stopped (after 10 CG steps) >>>>> with the following message: >>>>> >>>>> "*Sparse pattern is oversubscribed with nodes, please reduce number >>>>> of nodes*" >>>>> >>>>> The calculation are running using 7 and 3 cores (in the same node). In >>>>> both cases, the error message is the same. >>>>> >>>>> System: OpenSUSE 64bits, Intel and siesta-4.1--736 >>>>> >>>>> >>>>> I appreciate any help. >>>>> >>>>> []'s, >>>>> >>>>> Camps >>>>> >>>> >>>> >>>> -- >>>> Kind regards Nick >>>> >>> >> >> -- >> Kind regards Nick >> > -- Kind regards Nick
