"SUBPD",
	"SUBPS",
	"SUBQ",
	"SUBSD",
	"SUBSS",
	"SUBW",
	"SWAPGS",
	"SYSCALL",
	"SYSENTER",
	"SYSENTER64",
	"SYSEXIT",
	"SYSEXIT64",
	"SYSRET",
	"TESTB",
	"TESTL",
	"TESTQ",
	"TESTW",
	"TPAUSE",
	"TZCNTL",
	"TZCNTQ",
	"TZCNTW",
	"UCOMISD",
	"UCOMISS",
	"UD1",
	"UD2",
	"UMWAIT",
	"UNPCKHPD",
	"UNPCKHPS",
	"UNPCKLPD",
	"UNPCKLPS",
	"UMONITOR",

			v0.Aux = symToAux(sym)
			v0.AddArg2(ptr, mem)
			return true
		}
		break
	}
	// match: (TESTQ a:(ANDQload [off] {sym} x ptr mem) a)
	// cond: a.Uses == 2 && a.Block == v.Block && clobber(a)
	// result: (TESTQ (MOVQload <a.Type> [off] {sym} ptr mem) x)
	for {
		for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
			a := v_0
			if a.Op != OpAMD64ANDQload {

	LEAQ  256(inp), inp
	LEAQ  256(oup), oup
	SUBQ  $256, inl
	JMP   openSSEMainLoop

openSSEMainLoopDone:
	// Handle the various tail sizes efficiently
	TESTQ inl, inl
	JE    openSSEFinalize
	CMPQ  inl, $64
	JBE   openSSETail64
	CMPQ  inl, $128
	JBE   openSSETail128
	CMPQ  inl, $192
	JBE   openSSETail192
	JMP   openSSETail256

openSSEFinalize:

	MOVQ (8*2)(res_ptr), y_ptr
	MOVQ (8*3)(res_ptr), t1
	// Speculatively subtract
	SUBQ acc5, acc0
	SBBQ x_ptr, acc1
	SBBQ y_ptr, acc2
	SBBQ t1, acc3
	// If condition is 0, keep original value
	TESTQ t0, t0
	CMOVQEQ acc5, acc0
	CMOVQEQ x_ptr, acc1
	CMOVQEQ y_ptr, acc2
	CMOVQEQ t1, acc3
	// Store result
	MOVQ acc0, (8*0)(res_ptr)
	MOVQ acc1, (8*1)(res_ptr)
	MOVQ acc2, (8*2)(res_ptr)

	}

	if cursym.Func().Text.From.Sym.Wrapper() {
		// if g._panic != nil && g._panic.argp == FP {
		//   g._panic.argp = bottom-of-frame
		// }
		//
		//	MOVQ g_panic(g), regEntryTmp0
		//	TESTQ regEntryTmp0, regEntryTmp0
		//	JNE checkargp
		// end:
		//	NOP
		//  ... rest of function ...
		// checkargp:
		//	LEAQ (autoffset+8)(SP), regEntryTmp1
		//	CMPQ panic_argp(regEntryTmp0), regEntryTmp1

		// TESTx: compare (arg0 & arg1) to 0
		{name: "TESTQ", argLength: 2, reg: gp2flags, commutative: true, asm: "TESTQ", typ: "Flags"},
		{name: "TESTL", argLength: 2, reg: gp2flags, commutative: true, asm: "TESTL", typ: "Flags"},

//		}
//		// Use addr.
//		...
//	}
//
// nosplit because it is used during write barriers and must not be preempted.
//
//go:nosplit
func (tp typePointers) nextFast() (typePointers, uintptr) {
	// TESTQ/JEQ
	if tp.mask == 0 {
		return tp, 0
	}
	// BSFQ
	var i int
	if goarch.PtrSize == 8 {
		i = sys.TrailingZeros64(uint64(tp.mask))
	} else {
		i = sys.TrailingZeros32(uint32(tp.mask))
	}

		asm:            x86.ABTCQ,
		reg: regInfo{
			inputs: []inputInfo{
				{0, 4295032831}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 g R15 SB
			},
		},
	},
	{
		name:        "TESTQ",
		argLen:      2,
		commutative: true,
		asm:         x86.ATESTQ,
		reg: regInfo{
			inputs: []inputInfo{
				{0, 49151}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 R15