On OS X and iOS the following table shows you the the size of the types:

char       8 bit
short     16 bit
int       32 bit
long      system size
long long 64 bit

So, on 32bit OS X long is 32 bit, on 64bit OS X it’s 64 bit (unsure for iOS).

NSInteger has always the size of an pointer.

Often asked: “Use int or NSInteger?”. Answer: use NSInteger for pointers, ok usually you should use the pointer type itself, and for everything else use the size you require.

If you want exactly 8 bits, use int8_t for sigend or uint8_t for unsigned, for an integer which has at least those bits, use (u)int_least8_t (analogue for 16/32/64 bits).

Performance: you can use (u)int_fast8_t for the fastest type, which can hold (u)int_8_t. But since (u)int_leastY_t and (u)int_fastY_t are defined as (u)intY_t on OS X and iOS it should make no difference. I use the direct types for when all the bits are used, and the fast types when I don’t use exactly but less bits.

At last: don’t forget the LL or ULL suffix for 64bit constants (you can use (U)INT64_C(num) for always the correct suffix, also for 8/16/32), and the “%lld” in a formatted string for 64bit variables.
(example: NSLog(@"number: %lld", 1LL << 48);)

If you have more information or maybe corrections, please post a comment.