Divide by zero error in time_manager with ifx 2025.2.1 in debug mode

[DusanJovic-NOAA]

We are updating ufs-weather-model to spack-stack 2, which includes updating the Intel compiler (to version 2025.2.1) and FMS to 2025.03.

In some regression tests, we are getting a run-time floating-point exception error in FMS time_manager. This is the backtrace:

294: forrtl: error (73): floating divide by zero
294: Image              PC                Routine            Line        Source
294: libc.so.6          000014ECCD401BF0  Unknown               Unknown  Unknown
294: libimf.so          000014ECD8C048BD  __libm_log10_z0       Unknown  Unknown
294: fv3.exe            0000000008D08773  time_scalar_mult          778  time_manager.F90
294: fv3.exe            0000000008D08952  time_divide               806  time_manager.F90
294: fv3.exe            00000000055FC84D  initialize_mom           3529  MOM.F90
294: fv3.exe            0000000005502D92  ocean_model_init          286  mom_ocean_model_nuopc.F90
294: fv3.exe            000000000547B61F  initializeadverti         755  mom_cap.F90

See this ufs-weather-model issue for details: ufs-community/ufs-weather-model#3067

I created a small program that reproduces the error we are getting in the full ufs model. This only happens with the updated Intel compiler in a debug build mode. The release builds do not crash.

You can find the test program, on Gaea c6 in /gpfs/f6/drsa-hurr1/scratch/Dusan.Jovic/ufs/ss210/test_time_manager. Running ./build.sh should reproduce the error.

[laurenchilutti]

@DusanJovic-NOAA f6 will be unavailable beginning June 17 for 2 weeks. I would recommend putting your test program somewhere else, if possible.

[DusanJovic-NOAA]

@DusanJovic-NOAA f6 will be unavailable beginning June 17 for 2 weeks. I would recommend putting your test program somewhere else, if possible.

It appears that the floating-point error (divide by zero) can be resolved by reducing the compiler optimization for FMS (see the ufs-wm issue for details). We are still rebuilding FMS in spack-stack and running the ufs-wm regression tests. If the crash is indeed fixed and all tests pass, I'll close this issue.

[rickgrubin-noaa]

@DusanJovic-NOAA f6 will be unavailable beginning June 17 for 2 weeks. I would recommend putting your test program somewhere else, if possible.

@DusanJovic-NOAA is your reproducer code located elsewhere? I have a better solution to test. Thanks

[DusanJovic-NOAA]

@DusanJovic-NOAA f6 will be unavailable beginning June 17 for 2 weeks. I would recommend putting your test program somewhere else, if possible.

@DusanJovic-NOAA is your reproducer code located elsewhere? I have a better solution to test. Thanks

I think in ufs-wm issue in one of the comments you can find the test program.