(MOVVstore [0] ptr (MOVVconst [0]) mem)))

// medium zeroing uses a duff device
// 8, and 128 are magic constants, see runtime/mkduff.go
(Zero [s] {t} ptr mem)
	&& s%8 == 0 && s > 24 && s <= 8*128
	&& t.Alignment()%8 == 0 && !config.noDuffDevice =>
	(DUFFZERO [8 * (128 - s/8)] ptr mem)

// large or unaligned zeroing uses a loop
(Zero [s] {t} ptr mem)

	// Sort pointer-typed before non-pointer types.
	// Keeps the stack's GC bitmap compact.
	ap := a.Type().HasPointers()
	bp := b.Type().HasPointers()
	if ap != bp {
		return ap
	}

	// Group variables that need zeroing, so we can efficiently zero
	// them altogether.
	ap = a.Needzero()
	bp = b.Needzero()
	if ap != bp {
		return ap
	}

	// Sort variables in descending alignment order, so we can optimally

				(MOVDstore ptr (MOVDconst [0]) mem))))

// Medium 8-aligned zeroing uses a Duff's device
// 8 and 128 are magic constants, see runtime/mkduff.go
(Zero [s] {t} ptr mem)
	&& s%8 == 0 && s <= 8*128
	&& t.Alignment()%8 == 0 && !config.noDuffDevice =>
	(DUFFZERO [8 * (128 - s/8)] ptr mem)

// Generic zeroing uses a loop
(Zero [s] {t} ptr mem) =>
	(LoweredZero [t.Alignment()]
		ptr

	// not Linux decides to back this memory with transparent huge
	// pages. There's latency involved in this zeroing, but the hugepage
	// gains are almost always worth it. Note: it's important that we
	// clear even if it's freshly mapped and we know there's no point
	// to zeroing as *that* is the critical signal to use huge pages.
	memclrNoHeapPointers(unsafe.Pointer(s.base()), s.elemsize)
	s.needzero = 0

	b.Kind = kind
	b.ResetControls()
	b.Aux = nil
	b.AuxInt = 0
	b.Controls[0] = v
	b.Controls[1] = w
	v.Uses++
	w.Uses++
}

// truncateValues truncates b.Values at the ith element, zeroing subsequent elements.
// The values in b.Values after i must already have had their args reset,
// to maintain correct value uses counts.
func (b *Block) truncateValues(i int) {
	tail := b.Values[i:]

// already zeroed. Otherwise if needzero is true, objects are zeroed as
// they are allocated. There are various benefits to delaying zeroing
// this way:
//
//	1. Stack frame allocation can avoid zeroing altogether.
//
//	2. It exhibits better temporal locality, since the program is
//	   probably about to write to the memory.
//
//	3. We don't zero pages that never get reused.

             * corresponds to the least significant nonzero byte in lw ^ rw, since lw and rw are
             * little-endian. Long.numberOfTrailingZeros(diff) tells us the least significant
             * nonzero bit, and zeroing out the first three bits of L.nTZ gives us the shift to get
             * that least significant nonzero byte.
             */
            int n = Long.numberOfTrailingZeros(lw ^ rw) & ~0x7;

		{name: "LoweredZero", argLength: 2, reg: regInfo{inputs: []regMask{gp}}, aux: "Int64"},                    // large zeroing. arg0=start, arg1=mem, auxint=len, returns mem

		{name: "LoweredGetClosurePtr", reg: gp01},                                                                          // returns wasm.REG_CTXT, the closure pointer

             * corresponds to the least significant nonzero byte in lw ^ rw, since lw and rw are
             * little-endian. Long.numberOfTrailingZeros(diff) tells us the least significant
             * nonzero bit, and zeroing out the first three bits of L.nTZ gives us the shift to get
             * that least significant nonzero byte.
             */
            int n = Long.numberOfTrailingZeros(lw ^ rw) & ~0x7;

				clobbers: buildReg("R20 R21 R1"),
			},
			typ:            "Mem",
			faultOnNilArg0: true,
			faultOnNilArg1: true,
		},

		// large or unaligned zeroing
		// arg0 = address of memory to zero (in R20, changed as side effect)
		// arg1 = address of the last element to zero
		// arg2 = mem
		// auxint = alignment
		// returns mem
		//	MOVx	R0, (R20)