Question

I’m currently porting atomic.hpp out of boost for a project and would like to generalize the atomic add function whereby it’s templated on the type to add:

template <typename T, typename V>
inline T add(volatile T* mem, V val)
{
    T r;

    asm volatile
    (
        "lock\n\t"
        "xadd %1, %0":
        "+m"( *mem ), "=r"( r ):
        "1"( val ):
        "memory", "cc"
    );

    return r;
}

I can’t find a clear documentation as to wither it’s safe to use signed and unsigned 8, 16, 32 and 64 bit numbers with this. Anyone know?

Answer 1

Yes, you can use lock xadd under IA32 or IA64 with all types of numbers 8, 16, 32 and 64 bit signed or unsigned.
The memory alignment isn’t needed but if it is, then the memory access is faster.

From Intel Manual:

The integrity of the LOCK prefix is
not affected by the alignment of the
memory field. Memory locking is
observed for arbitrarily misaligned
fields. This instruction’s operation
is the same in non-64-bit modes and
64-bit mode.

Warning!:

If the LOCK prefix is used with XADD instruction and the source
operand is a memory operand, an
undefined opcode exception (#UD) may
be generated.

So the source operand must be register and the destination operand is memory address!