(MOVBQZX x:(MOVBload [off] {sym} ptr mem)) && x.Uses == 1 && clobber(x) => @x.Block (MOVBload <v.Type> [off] {sym} ptr mem)
(MOVBQZX x:(MOVWload [off] {sym} ptr mem)) && x.Uses == 1 && clobber(x) => @x.Block (MOVBload <v.Type> [off] {sym} ptr mem)
(MOVBQZX x:(MOVLload [off] {sym} ptr mem)) && x.Uses == 1 && clobber(x) => @x.Block (MOVBload <v.Type> [off] {sym} ptr mem)

			}

			// Dump any registers which will be clobbered
			if s.doClobber && v.Op.IsCall() {
				// clobber registers that are marked as clobber in regmask, but
				// don't clobber inputs.
				s.clobberRegs(regspec.clobbers &^ s.tmpused &^ s.nospill)
			}
			s.freeRegs(regspec.clobbers)
			s.tmpused |= regspec.clobbers

			// Pick registers for outputs.
			{

  && clobber(x)
  => (STMG3 [i-16] {s} p w0 w1 w2 mem)
(MOVDstore [i] {s} p w3 x:(STMG3 [i-24] {s} p w0 w1 w2 mem))
  && x.Uses == 1
  && is20Bit(int64(i)-24)
  && setPos(v, x.Pos)
  && clobber(x)
  => (STMG4 [i-24] {s} p w0 w1 w2 w3 mem)
(STMG2 [i] {s} p w2 w3 x:(STMG2 [i-16] {s} p w0 w1 mem))
  && x.Uses == 1
  && is20Bit(int64(i)-16)
  && setPos(v, x.Pos)
  && clobber(x)
  => (STMG4 [i-16] {s} p w0 w1 w2 w3 mem)

// gcWriteBarrier does NOT follow the Go ABI. It accepts the
// number of bytes of buffer needed in R11, and returns a pointer
// to the buffer space in R11.
// It clobbers FLAGS. It does not clobber any general-purpose registers,
// but may clobber others (e.g., SSE registers).
// Typical use would be, when doing *(CX+88) = AX
//     CMPL    $0, runtime.writeBarrier(SB)
//     JEQ     dowrite

(MFVSRD (FMOVDconst [c])) => (MOVDconst [int64(math.Float64bits(c))])

(MTVSRD x:(MOVDload [off] {sym} ptr mem)) && x.Uses == 1 && clobber(x) => @x.Block (FMOVDload [off] {sym} ptr mem)
(MFVSRD x:(FMOVDload [off] {sym} ptr mem)) && x.Uses == 1 && clobber(x) => @x.Block (MOVDload [off] {sym} ptr mem)

// Rules for MOV* or FMOV* ops determine when indexed (MOV*loadidx or MOV*storeidx)

	return nil // too far away
}

// clobber invalidates values. Returns true.
// clobber is used by rewrite rules to:
//
//	A) make sure the values are really dead and never used again.
//	B) decrement use counts of the values' args.
func clobber(vv ...*Value) bool {
	for _, v := range vv {
		v.reset(OpInvalid)
		// Note: leave v.Block intact.  The Block field is used after clobber.
	}
	return true
}

		// It saves all GP registers if necessary,
		// but clobbers R14 (LR) because it's a call,
		// and also clobbers R1 as the PLT stub does.
		// Returns a pointer to a write barrier buffer in R9.
		{name: "LoweredWB", argLength: 1, reg: regInfo{clobbers: (callerSave &^ gpg) | buildReg("R14") | r1, outputs: []regMask{r9}}, clobberFlags: true, aux: "Int64"},

		// It saves all GP registers if necessary,
		// but clobbers R30 (LR) because it's a call.
		// R16 and R17 may be clobbered by linker trampoline.
		// Returns a pointer to a write barrier buffer in R25.
		{name: "LoweredWB", argLength: 1, reg: regInfo{clobbers: (callerSave &^ gpg) | buildReg("R16 R17 R30"), outputs: []regMask{buildReg("R25")}}, clobberFlags: true, aux: "Int64"},

	}

	// Handle any register clobbering. Call operations, for example,
	// clobber all registers even though they don't explicitly write to
	// them.
	clobbers := uint64(opcodeTable[v.Op].reg.clobbers)
	for {
		if clobbers == 0 {
			break
		}
		reg := uint8(bits.TrailingZeros64(clobbers))
		clobbers &^= 1 << reg

		for _, slot := range locs.registers[reg] {
			if state.loggingLevel > 1 {

	LXVD2X (R16)(CPOOL), P1
	LXVD2X (R0)(CPOOL), P0

	CALL p256MulInternal<>(SB)

	MOVD	n+16(FP), N
	ADD	$-1, N
	CMP	$0, N
	BEQ	done
	MOVD	N, n+16(FP)	// Save counter to avoid clobber
	VOR	T0, T0, X0
	VOR	T1, T1, X1
	BR	sqrLoop

done:
	MOVD $p256mul<>+0x00(SB), CPOOL

	XXPERMDI T0, T0, $2, T0
	XXPERMDI T1, T1, $2, T1
	STXVD2X T0, (R0)(res_ptr)