Results are garbage when compiling with ifx and -O2

[elinscott]

kcp.x compiles but gives garbage when compiled with ifx. To be more specific, the calculation runs to completion but quantities such as total energy, kinetic energy, electrostatic energy, HOMO/LUMO are wrong (not orders-of-magnitude wrong, "just" off by several/many Hartrees). Meanwhile, esr and eself are exactly the same as results from ifort.

So far I have discovered that

• compiling with O1 fixes it
• compiling with -check bounds fixes it (why, I do not know)
• compiling with an older version of oneapi does not fix it (I have tried 2024.0, 2024.1, 2024.2, and 2025.0)

I will continue to investigate...