https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99151
--- Comment #7 from Eric Botcazou <ebotcazou at gcc dot gnu.org> --- > This nop behaviour could be a bit inconsistent across architectures. For > example, arm and powerpc don't generate a nop here. Well, it's low-level trickery so architecture-dependent by definition. > I still think that the profiling counter increment in the > __builtin_unreachable() path is a bug. How so? I only see a missed optimization, but with -fprofile-arcs -ftest-coverage you're splitting hairs IMO.