amrex::Abort::0:: (-1,0,0,0) is out of bound (0:63,0:63,0:31,0:0)

[wang1202]

Hello, I'm testing the FlowInABox case, and I found that when I apply grid refinement, the following error message appears when producing the next level:

amrex::Abort::0:: (-1,0,0,0) is out of bound (0:63,0:63,0:31,0:0) !!!

Using the debug function in VSCode, I noticed that the bounds include 3 ghost points added to the boundaries, but it seems these points are not added to the higher level. Where can I check whether the ghost points are being added to the higher levels? Below are what I tested in the input file:

REFINEMENT / REGRIDDING

amr.max_level = 1 # maximum level number allowed
amr.compute_new_dt_on_regrid = true

erf.coupling_type = "TwoWay"
erf.regrid_int = 1
erf.refinement_indicators = diff_theta

erf.diff_theta.max_level = 1
erf.diff_theta.adjacent_difference_greater = 1.
erf.diff_theta.field_name = theta
erf.diff_theta.start_time = 10.0

[asalmgren]

Hey @wang1202 -- thanks for pointing this out! Let me take a look at this tonight and get back to you -- I haven't tested this problem recently ...

[asalmgren]

@wang1202 -- I believe this issue is fixed by PR 2055 -- could you give that a try? Please let me know if it works.

[wang1202]

Hi @asalmgren -- thank you for the quick reply. After updating the code and running again, the out-of-bounds error seems to be gone. However, the segmentation fault still occurs. Below is the output where the run stops:

"""
Coarse STEP 5 starts ...
Based on cfl of 1.0
Compressible dt at level 0 would be: 9.232827232e-05
Anelastic dt at level 0 would be: 172.6188152
Fixed dt at level 0 is: 0.025
Creating new distribution map on level: 1
NEW BA FROM COARSE AT LEVEL 1 (BoxArray maxbox(2)
m_ref->m_hash_sig(0)
((0,0,0) (127,127,1) (0,0,0)) ((0,0,62) (127,127,63) (0,0,0)) )

Segfault
See Backtrace.0 file for details
"""

And below is the content of Backtrace.0:
"""
0: amrex::BLBackTrace::print_backtrace_info(__sFILE*) (in ERF3d.gnu.DEBUG.MPI.ex) (AMReX_BLBackTrace.cpp:203)

1: amrex::BLBackTrace::handler(int) (in ERF3d.gnu.DEBUG.MPI.ex) (AMReX_BLBackTrace.cpp:99)

2: _sigtramp (in libsystem_platform.dylib) + 56

3: amrex::Geometry::isAllPeriodic() const (in ERF3d.gnu.DEBUG.MPI.ex) (AMReX_Geometry.H:340)

4: ERFPhysBCFunct_base::operator()(amrex::MultiFab&, int, int, amrex::IntVectND<3> const&) (in ERF3d.gnu.DEBUG.MPI.ex) (ERF_PhysBCFunct.cpp:315)

5: ERF::initHSE(int) (in ERF3d.gnu.DEBUG.MPI.ex) (ERF_Init1D.cpp:65)

6: ERF::MakeNewLevelFromCoarse(int, double, amrex::BoxArray const&, amrex::DistributionMapping const&) (in ERF3d.gnu.DEBUG.MPI.ex) (ERF_MakeNewLevel.cpp:284)

7: amrex::AmrCore::regrid(int, double, bool) (in ERF3d.gnu.DEBUG.MPI.ex) (AMReX_AmrCore.cpp:121)

8: ERF::timeStep(int, double, int) (in ERF3d.gnu.DEBUG.MPI.ex) (ERF_TimeStep.cpp:55)

9: ERF::Evolve() (in ERF3d.gnu.DEBUG.MPI.ex) (ERF.cpp:398)

10: main (in ERF3d.gnu.DEBUG.MPI.ex) (main.cpp:128)

11: start (in dyld) + 2840
"""

[asalmgren]

@wang1202 -- sorry not to fix this sooner -- somehow in my head this was resolved but it definitely wasn't. I do believe that PR 2077 fixes this -- I am running with your multilevel inputs and have definitely gotten past the error which occurred when first creating the level 1 grids. I'll close this now but please re-open if there are any more problems!