search

Nek5000 / nekRS

our next generation fast and scalable CFD code
https://nek5000.mcs.anl.gov/
Other
284 stars 75 forks source link

Simulation crashes at same time step for one supercomputer but runs successfully on other #479

Closed Platinumd1991 closed 1 year ago

Platinumd1991 commented 1 year ago

Hi,

I don;t know why my simulation crashes on one supercomputer (GPU) but not the other (CPU). I have tried to reduce delta t to lower the courant number but my simulation still crashes as the same exact timestep as posted below when I use (NekRS) but this same case runs successfully on Nek5000. What could be the cause of this? My collegues are also having the same issue with their cases between the twosupercomputers.

step= 52598 t= 1.82207840e+02 dt=8.0e-05 C= 4.27 UVW: 2 P: 22 eTime= 2.11e-01, 2.71779e+03 s step= 52599 t= 1.82207920e+02 dt=8.0e-05 C= 5.82 UVW: 3 P: 23 eTime= 2.09e-01, 2.71800e+03 s step= 52600 t= 1.82208000e+02 dt=8.0e-05 C= 8.76 UVW: 3 P: 25 eTime= 2.30e-01, 2.71823e+03 s copying solution to nek calling nek_userchk ... 52600 1.82208000000E+02 7.26226313888E+00 7.26559569350E+00 -3.33255462393E-03 1dragx 52600 1.82208000000E+02 -2.88376652997E-03 -2.62383438671E-03 -2.59932143255E-04 1dragy 52600 1.82208000000E+02 -2.16552739071E-04 1.40683273030E-16 -2.16552739071E-04 1dragz 52600 1.82208000000E+02 7.26226313888E+00 7.26559569350E+00 -3.33255462393E-03 2dragx 52600 1.82208000000E+02 -2.88376652997E-03 -2.62383438671E-03 -2.59932143255E-04 2dragy 52600 1.82208000000E+02 -2.16552739071E-04 1.40683273030E-16 -2.16552739071E-04 2dragz dump history points step= 52601 t= 1.82208080e+02 dt=8.0e-05 C= 15.38 UVW: 3 P: 30 eTime= 3.06e-01, 2.71854e+03 s step= 52602 t= 1.82208160e+02 dt=8.0e-05 C= 45.05 UVW: 3 P: 37 eTime= 3.26e-01, 2.71886e+03 s Unreasonable CFL! Dying ...

stgeke commented 1 year ago

Please use https://github.com/Nek5000/nekRS/discussions for things like this.