https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115896
--- Comment #1 from Gabriel Ravier <gabravier at gmail dot com> ---
A generalization of this is that `(a << cst0) == cst1` should be foldable to `a
== cst0 >> cst1`, leading to the much more obvious case of:
bool f1(int8_t a)
{
int32_t b = a << 4;
return b == -128;
}
being optimized by LLVM to `a == -8` (I think since commit
291c3d8ff21652a7bd654cfc170e0891e46a84da), which GCC does not do.
