MOVL	delta_lo+4(FP), SI
	MOVL	delta_hi+8(FP), DI
	// DX:AX = *addr
	MOVL	0(BP), AX
	MOVL	4(BP), DX
addloop:
	// CX:BX = DX:AX (*addr) + DI:SI (delta)
	MOVL	AX, BX
	MOVL	DX, CX
	ADDL	SI, BX
	ADCL	DI, CX

	// if *addr == DX:AX {
	//	*addr = CX:BX
	// } else {
	//	DX:AX = *addr
	// }
	// all in one instruction
	LOCK
	CMPXCHG8B	0(BP)

	JNZ	addloop

	// success

	// be inlinable. If they have no callers in their packages, they
	// might not actually be inlined anywhere.
	want := map[string][]string{
		"runtime": {
			"add",
			"acquirem",
			"add1",
			"addb",
			"adjustpanics",
			"adjustpointer",
			"alignDown",
			"alignUp",
			"bucketMask",
			"bucketShift",
			"chanbuf",
			"evacuated",
			"fastlog2",
			"float64bits",
			"funcspdelta",
			"getm",

    <inspection_tool class="21f28a86-2ba8-3048-9ce1-b6f10d4d4a12" enabled="true" level="WEAK WARNING" enabled_by_default="true" editorAttributes="INFO_ATTRIBUTES" />
    <inspection_tool class="73cd39e8-e54e-3a38-ad1b-b883fff4b1eb" enabled="true" level="WEAK WARNING" enabled_by_default="true" editorAttributes="INFO_ATTRIBUTES" />
    <inspection_tool class="AsciiDocLinkResolve" enabled="false" level="ERROR" enabled_by_default="false" />

		Load(aX, RAX)
	default:
		Comment(fmt.Sprintf("%s += %d×%s×%s", r, i, aX, bX))
		IMUL3Q(Imm(i), Load(aX, GP64()), RAX)
	}
	MULQ(mustAddr(bX)) // RDX, RAX = RAX * bX
	ADDQ(RAX, r.lo)
	ADCQ(RDX, r.hi)
}

// shiftRightBy51 returns r >> 51 and r.lo.
//
// After this function is called, the uint128 may not be used anymore.
func shiftRightBy51(r *uint128) (out, lo GPVirtual) {
	out = r.hi

	// due to one of the throws below shows the original block
	// base and extent.
	b := b0
	n := n0

	for i := uintptr(0); i < n; {
		// Find bits for the next word.
		bits := uint32(*addb(ptrmask, i/(goarch.PtrSize*8)))
		if bits == 0 {
			i += goarch.PtrSize * 8
			continue
		}
		for j := 0; j < 8 && i < n; j++ {
			if bits&1 != 0 {
				// Same work as in scanobject; see comments there.

	p := typ.GCData // start of 1-bit pointer mask (or GC program)
	var gcProgBits uintptr
	if typ.Kind_&abi.KindGCProg != 0 {
		// Expand gc program, using the object itself for storage.
		gcProgBits = runGCProg(addb(p, 4), (*byte)(ptr))
		p = (*byte)(ptr)
	}
	nb := typ.PtrBytes / goarch.PtrSize

	for i := uintptr(0); i < nb; i += ptrBits {
		k := nb - i
		if k > ptrBits {
			k = ptrBits
		}

	{ALDREX, C_SOREG, C_NONE, C_REG, 77, 4, 0, 0, 0, 0},
	{ASTREX, C_SOREG, C_REG, C_REG, 78, 4, 0, 0, 0, 0},
	{ADMB, C_NONE, C_NONE, C_NONE, 110, 4, 0, 0, 0, 0},
	{ADMB, C_LCON, C_NONE, C_NONE, 110, 4, 0, 0, 0, 0},
	{ADMB, C_SPR, C_NONE, C_NONE, 110, 4, 0, 0, 0, 0},
	{AMOVF, C_ZFCON, C_NONE, C_FREG, 80, 8, 0, 0, 0, 0},
	{AMOVF, C_SFCON, C_NONE, C_FREG, 81, 4, 0, 0, 0, 0},

			oprangeset(AUCVTFWD, t)
			oprangeset(AUCVTFWS, t)

		case ASYS:
			oprangeset(AAT, t)
			oprangeset(AIC, t)

		case ATLBI:
			oprangeset(ADC, t)

		case ASYSL, AHINT:
			break

		case ADMB:
			oprangeset(ADSB, t)
			oprangeset(AISB, t)

		case AMRS, AMSR:
			break

		case ALDAR:
			oprangeset(ALDARW, t)
			oprangeset(ALDARB, t)

var unaryDst = map[obj.As]bool{
	ABSWAPL:     true,
	ABSWAPQ:     true,
	ACLDEMOTE:   true,
	ACLFLUSH:    true,
	ACLFLUSHOPT: true,
	ACLWB:       true,
	ACMPXCHG16B: true,
	ACMPXCHG8B:  true,
	ADECB:       true,
	ADECL:       true,
	ADECQ:       true,
	ADECW:       true,
	AFBSTP:      true,
	AFFREE:      true,
	AFLDENV:     true,
	AFSAVE:      true,
	AFSTCW:      true,
	AFSTENV:     true,

[...]Optab{
	{obj.AXXX, nil, 0, opBytes{}},
	{AAAA, ynone, P32, opBytes{0x37}},
	{AAAD, ynone, P32, opBytes{0xd5, 0x0a}},
	{AAAM, ynone, P32, opBytes{0xd4, 0x0a}},
	{AAAS, ynone, P32, opBytes{0x3f}},
	{AADCB, yxorb, Pb, opBytes{0x14, 0x80, 02, 0x10, 0x12}},
	{AADCL, yaddl, Px, opBytes{0x83, 02, 0x15, 0x81, 02, 0x11, 0x13}},
	{AADCQ, yaddl, Pw, opBytes{0x83, 02, 0x15, 0x81, 02, 0x11, 0x13}},