[FLASH-USERS] Hydro timestep calculation and paramesh error

[Charles Heaton]

Hi,

I've been running Flash on SCARF and receive two errors that seem to be persistent with the current batch of simulations I'm trying (I've done very similar simulations which haven't caused these problems so I don't understand what's going wrong).


  1.
Hydro timestep calculation:
     *
The hydro timestep for some reason seems to drop very quickly, see attached figure, to values of ~ 1E-19s. I don't understand the logic behind this. I'm running with a maximum refinement that gives min(dx) ~ 0.2 micron. A timestep of this magnitude implies some process propagating at speeds greater than the speed of light (my CFL multiplier is only 0.1). The minimum timestep of the simulation is 1E-16 and so the simulation often fails after due to negative 3T temperatures (which I assume is coming from time stepping at an interval much larger than whatever hydro timestep is calculated).
     *
I receive warnings at every timestep

[dR1St]+ fallback to order 1 for iDim=           2 at ix,iy=          15
          11  in Block          86 @           3
 [dR1St]+   5395063991.13334        894594893.143532
   256288896.149746        23571715.8670185       9.595549910113477E-006
 -2.918341007477648E-006

            
          These errors are written many many times per timestep which render the standard output unusable without filtering with                                 regular expressions.


  1.
 Paramesh error:
     *
This I suspect is a SCARF dependant issue, but I'd be interested to see what people think. Whenever I run on more than one node (24 cores on each node), I receive (usually very shortly after the program has started) an AMR error that says:

[amr_prolong_gen_unk1_fun] PE=     25, ivar=  5, value=1.248E+04
 Trying to convert non-zero mass-specific variable to per-volume form in monoton
 ic mesh interpolation, but dens is zero!
 DRIVER_ABORT:
 Trying to convert non-zero mass-specific variable to per-volume form, but dens
 is zero!
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 25


This is probably something to do with the way the simulation is setup and compiled so for clarity, I've loaded the
  2.
            modules included below and the Makefile is attached. I'm using -2d +cylindrical -maxblocks=1000 +pm4dev -nxb=16                            -nyb=16 flags in the setup command.
     *

     *
module load visit
     *
module load intel/17.0.2
     *
module load intel/mkl/17.0.2
     *
module load Python/2.7
     *
module load intel/mpi/17.0.2
     *
module load libPHDF5 # Version is 1.10.1-intel-17.0.2

Any help with this will be greatly appreciated.

Thanks,
Charlie
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20240313/d8385c52/attachment.htm>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: Makefile.h
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20240313/d8385c52/attachment.h>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Timesteps.jpg
Type: image/jpeg
Size: 69347 bytes
Desc: Timesteps.jpg
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20240313/d8385c52/attachment.jpg>