I support a large CFD code and I've had some job failures using -O2 for ifx version 2025.1 and 2025.2. These jobs run successfully for 2025.0.4. The jobs also run successfully when using -O1 of the 2025.1 or 2025.2. So I believe that there is something about the -O2 option in the latest release. The vast majority of my test cases still run fine with the latest version, so this is not happening for all cases. My question is -- is there some way that I might diagnose the problem? I have added -g -traceback to the compilation arguments, but this only led to a line number that was the last line of a source code file, the end module statement. So I don't know where there might be a problem, or where the numerics of the -O1 case diverge from the -O2 case.
