IBM GPU run fails when viscosity is disabled

[fahnab666]

Describe the bug
GPU IBM inviscid simulation fails during simulation for a 3D single-sphere case that should be valid. The setup is an inviscid Euler case (viscous = F) with one fixed IBM sphere and a shock initialized using pre-shock/post-shock patches. The run passes setup stages but crashes once the GPU simulation starts.

To Reproduce

1. Compile MFC with GPU support on Tuolumne/AMD GPUs.
2. Run the attached inviscid IBM reproducer case (case.py) on GPU.
3. syscheck passes, pre_process completes, and simulation aborts.

Example job flow:

1. Run syscheck
2. Run pre_process
3. Run simulation
4. Observe GPU-side abort / memory access fault

Representative error:

Memory access fault by GPU ... on address (nil). Reason: Unknown.
Failed to allocate file: Bad file descriptor
GPU core dump failed
...
mfc: ERROR > :( .../bin/simulation failed with exit code 134.


Expected behavior
This inviscid IBM slip-wall single-sphere case should run on GPU without crashing. Since syscheck passes and pre_process completes successfully, simulation is expected to advance normally instead of aborting with GPU memory access faults.

Screenshots
N/A

Environment

System: LLNL Tuolumne

GPUs: AMD

Run mode: GPU

Nodes: 2

MPI ranks: 8

GPUs per task: 1

Relevant environment: Tuo gpu environment.

Additional context
with high Re, it runs, just not with the viscosity turn off. 


added file  :
[case.py](https://github.com/user-attachments/files/25830803/case.py)

[invis-sim.txt](https://github.com/user-attachments/files/25830826/invis-sim.txt) its the output file


[invis-sim_error.txt](https://github.com/user-attachments/files/25830848/invis-sim_error.txt)

 
Possible issue:
GPU IBM handling for inviscid cases may be dereferencing a null pointer or hitting an invalid device memory path during simulation, since the crash happens after preprocessing and reports GPU memory access fault on (nil).

[sbryngelson]

Did you try running on any other GPU system? This is almost surely Tuo specific.

[fahnab666]

I haven't, But I will run it to Hipergator and Update here tomorrow. Best, Fahad