upper 16 bits are usually unused on x86_64 as well (since the x86_64 virtual address space is only 48-bit), though whether they're one or zero depends on bit 48
And doubles have some 2^50 NaN's. It's possible to create a tagged pointer of either a double, 32-bit integer, pointer or a small string and still have space for more.
This reminds me of all the problems Apple had with 24 bit addressing and smuggling data in the unused eight bits at the top of each pointer. Andy Hertzfeld considers that his worst blunder:
Not UB but rather implementation defined. The compiler will not do anything weird here, but you rely on platform behavior that's not specified by the C lang spec.
Not UB in the slightest. You can take a pointer, cast it to uintptr_t, do whatever you want to that int, then cast it back to a pointer. It's only UB if the resulting pointer isn't valid or you cast it to the wrong type (and strict aliasing is enabled.
