double-word compare-and-swap function using cmpxchg16b (D language)

dwcas16.d

// by Niklas Schulze / nischu7 <n7@pronghorn.org>
//
// requires cmpxchg16b CPU instruction
// required GDC patch: http://bitbucket.org/nischu7/gdc/changeset/d8a2a73fb3d8
bool dwcas16(T, V1, V2)(T* here, const V1 ifThis, const V2 writeThis)
{
    asm
    {
        /*
        mov RAX, ifThis;       // copy ifThis[0] to RAX
        mov RDX, 8[ifThis];    // copy ifThis[1] to RDX

        mov RBX, writeThis;    // copy writeThis[0] to RBX
        mov RCX, 8[writeThis]; // copy writeThis[1] to RCX
        */

        // slightly faster?
        lea RSI, ifThis;
        mov RAX, [RSI];
        mov RDX, 8[RSI];

        lea RSI, writeThis;
        mov RBX, [RSI];
        mov RCX, 8[RSI];

        mov RSI, here;

        lock; // lock always needed to make this op atomic
        cmpxchg16b [RSI];

        setz AL;
    }
}